Computerworld.com [Hacking News]

Syndikovat obsah
Making technology work for business
Aktualizace: 20 min 53 sek zpět

Adobe Firefly expands with ‘commercially safe’ video generator

13 Únor, 2025 - 19:20

Adobe has released a video generator in public beta in its generative AI (genAI) tool, Adobe Firefly. The company calls the tool the first “commercially safe” video generator on the market. It has been trained on licensed content and public domain material, meaning it should not be able to generate material that could infringe someone else’s copyright.

Firefly can generate clips either from text instructions or by combining a reference image with text instructions. There are also settings to customize things such as camera angles, movements, and distances.

A paid subscription is required to use the video generator. Firefly Standard, which costs about $11 a month, gives access to 2000 credits; that should be enough for 20 five-second videos with a 1080p picture resolution and a frame rate of 24 frames per second.

Firefly Pro, which costs three times more than the standard version, allows a user 7000 credits, which should be enough for 70 five-second clips in 1080p at 24 frames per second.

Adobe plans to eventually release a model for videos with lower resolution but faster image updates, as well as a model with 4k resolution for Pro users.

Kategorie: Hacking & Security

AI chatbots outperform doctors in diagnosing patients, study finds

13 Únor, 2025 - 12:00

Chatbots quickly surpassed human physicians in diagnostic reasoning — the crucial first step in clinical care — according to a new study published in the journal Nature Medicine.

The study suggests physicians who have access to large language models (LLMs), which underpin generative AI (genAI) chatbots, demonstrate improved performance on several patient care tasks compared to colleagues without access to the technology.

The study also found that physicians using chatbots spent more time on patient cases and made safer decisions than those without access to the genAI tools.

The research, undertaken by more than a dozen physicians at Beth Israel Deaconess Medical Center (BIDMC), showed genAI has promise as an “open-ended decision-making” physician partner.

“However, this will require rigorous validation to realize LLMs’ potential for enhancing patient care,” said Dr. Adam Rodman, director of AI Programs at BIDMC. “Unlike diagnostic reasoning, a task often with a single right answer, which LLMs excel at, management reasoning may have no right answer and involves weighing trade-offs between inherently risky courses of action.”

The conclusions were based on evaluations about the decision-making capabilities of 92 physicians as they worked through five hypothetical patient cases. They focused on the physicians’ management reasoning, which includes decisions on testing, treatment, patient preferences, social factors, costs, and risks.

When responses to their hypothetical patient cases were scored, the physicians using a chatbot scored significantly higher than those using conventional resources only. Chatbot users also spent more time per case — by nearly two minutes — and they had a lower risk of mild-to-moderate harm compared to those using conventional resources (3.7% vs. 5.3%). Severe harm ratings, however, were similar between groups.

“My theory,” Rodman said, “[is] the AI improved management reasoning in patient communication and patient factors domains; it did not affect things like recognizing complications or medication decisions. We used a high standard for harm — immediate harm — and poor communication is unlikely to cause immediate harm.”

An earlier 2023 study by Rodman and his colleagues yielded promising, yet cautious, conclusions about the role of genAI technology. They found it was “capable of showing the equivalent or better reasoning than people throughout the evolution of clinical case.”

That data, published in Journal of the American Medical Association (JAMA), used a common testing tool used to assess physicians’ clinical reasoning. The researchers recruited 21 attending physicians and 18 residents, who worked through 20 archived (not new) clinical cases in four stages of diagnostic reasoning, writing and justifying their differential diagnoses at each stage.

The researchers then performed the same tests using ChatGPT based on the GPT-4 LLM. The chatbot followed the same instructions and used the same clinical cases. The results were both promising and concerning.

The chatbot scored highest in some measures on the testing tool, with a median score of 10/10, compared to 9/10 for attending physicians and 8/10 for residents. While diagnostic accuracy and reasoning were similar between humans and the bot, the chatbot had more instances of incorrect reasoning. “This highlights that AI is likely best used to augment, not replace, human reasoning,” the study concluded.

Simply put, in some cases “the bots were also just plain wrong,” the report said.

Rodman said he isn’t sure why the genAI study pointed to more errors in the earlier study. “The checkpoint is different [in the new study], so hallucinations might have improved, but they also vary by task,” he said. “ Our original study focused on diagnostic reasoning, a classification task with clear right and wrong answers. Management reasoning, on the other hand, is highly context-specific and has a range of acceptable answers.”

A key difference from the original study is the researchers are now comparing two groups of humans — one using AI and one not — while the original work compared AI to humans directly. “We did collect a small AI-only baseline, but the comparison was done with a multi-effects model. So, in this case, everything is mediated through people,” Rodman said.

Researcher and lead study author Dr. Stephanie Cabral, a third-year internal medicine resident at BIDMC, said more research is needed on how LLMs can fit into clinical practice, “but they could already serve as a useful checkpoint to prevent oversight.

“My ultimate hope is that AI will improve the patient-physician interaction by reducing some of the inefficiencies we currently have and allow us to focus more on the conversation we’re having with our patients,” she said.

The latest study involved a newer, upgraded version of GPT-4, which could explain some of the variations in results.

To date, AI in healthcare has mainly focused on tasks such as portal messaging, according to Rodman. But chatbots could enhance human decision-making, especially in complex tasks.

“Our findings show promise, but rigorous validation is needed to fully unlock their potential for improving patient care,” he said. “This suggests a future use for LLMs as a helpful adjunct to clinical judgment. Further exploration into whether the LLM is merely encouraging users to slow down and reflect more deeply, or whether it is actively augmenting the reasoning process would be valuable.”

The chatbot testing will now enter the next of two follow-on phases, the first of which has already produced new raw data to be analyzed by the researchers, Rodman said. The researchers will begin looking at varying user interaction, where they study different types of chatbots, different user interfaces, and doctor education about using LLMs (such as more specific prompt design) in controlled environments to see how performance is affected.

The second phase will also involve real-time patient data, not archived patient cases.

“We are also studying [human computer interaction] using secure LLMs — so [it’s] HIPAA complaint — to see how these effects hold in the real world,” he said.

Kategorie: Hacking & Security

OpenAI revamps AI roadmap, merging models for a leaner future

13 Únor, 2025 - 10:48

OpenAI will integrate “o3” into GPT-5 instead of releasing it separately, streamlining adoption while signaling a shift toward fewer, more controlled AI models amid rising competition and cost pressures.

“In both ChatGPT and our API, we will release GPT-5 as a system that integrates a lot of our technology, including o3,” CEO Sam Altman said in a post on X.

The decision marks a departure from OpenAI’s recent strategy of offering multiple model variants, suggesting the company is prioritizing ease of deployment and product clarity for enterprise users.

“We want AI to ‘just work’ for you; we realize how complicated our model and product offerings have gotten,” Altman said. “We hate the model picker as much as you do and want to return to magic unified intelligence.”

With enterprises facing rising costs for AI adoption and competitors like DeepSeek introducing lower-cost alternatives, OpenAI’s move could also be a response to market pressures.

A single, more comprehensive model may help justify AI investments by reducing the complexity of integrating multiple systems while ensuring compatibility with OpenAI’s broader ecosystem.

OpenAI will also launch GPT-4.5, codenamed “Orion,” as its final model without chain-of-thought reasoning, Altman added, without providing a timeline.

A change of approach


The rapid proliferation of AI models has intensified competition among research labs, each striving to develop smarter, more efficient systems with larger context windows and specialized functions.

While this innovation has expanded capabilities, it has also introduced complexity, making it harder for users to choose the right model.

“The burgeoning list of models has added complexity for the average user who just wants chat to work without having to figure out which model to use,” said Abhishek Sengupta, practice director at Everest Group. “For developers, it’s a mixed bag – on one hand it takes away the need to incessantly check which model is best suited for which task (at least for OpenAI) but on the other hand you are outsourcing your choice of optimal model to OpenAI.”

While model selection may still occur, OpenAI could handle the process rather than users. Analysts suggest this could also be an attempt to avoid the race between model performance and cost by bundling all AI capabilities under a single system.

“Maybe the consolidation of models into a single source of intelligence is a move toward creating an intelligence platform,” Sengupta added. “Maybe that’s the differentiation they are placing their bets on. Time will tell.”

Rising competition and open-source threats

This shift could also reshape the economics of AI, giving OpenAI greater control over costs, deployment, and market positioning.

“I believe merging it has multiple benefits, not just in terms of costs related to training, go-to-market strategies, and customer delivery, but also in giving OpenAI more leverage to drive it as a ‘system’ and extract more value through a simplified business model,” said Neil Shah, partner and co-founder at Counterpoint Research. “This will change the economics on both ends, which investors will be keen to monitor and measure.”

This comes at a time when AI competition is intensifying, with DeepSeek disrupting the market with cost-effective models, highlighting the pressure on OpenAI to refine its strategy.

“One cannot rule out this move being triggered by competitive models like DeepSeek, which are highly cost-effective,” said Sanchit Vir Gogia, chief analyst and CEO at Greyhound Research. “Of course, there shall be many other models out there that will be more cost-effective and innovative, and most importantly, will be made open source and not proprietary like OpenAI.”

Importantly, not all organizations have the resources, need, or strategic planning to navigate complex, tiered pricing structures. “Despite the rise of SaaS, many large enterprises prefer EULA contracts since they are incubated from any risk associated with sudden and unplanned need for resources,” Gogia added. “In the same breath, not all organizations require a customized model and the flexibility that comes along with it. Many of their use cases are simplistic enough to use a model that keeps the billing and the use simple.”

Kategorie: Hacking & Security

EU pulls back – for the moment – on privacy and genAI liability compliance regulations

13 Únor, 2025 - 04:44

When the EU on Tuesday said it was not, at this time, moving ahead with critical legislation involving privacy and genAI liability issues, it honestly reported that members couldn’t agree. But the reasons why they couldn’t agree get much more complicated.

The EU decisions involved two seemingly unrelated pieces of legislation: One dealing with privacy efforts, often called the cookie law, and the other dealing with AI liability. 

The EU decisions are in the annexes to the Commission’s  work programme for 2025, in Annex IV, items 29 and 32. For the AI liability section (“on adapting non-contractual civil liability rules to artificial intelligence”), the EU found “no foreseeable agreement. The Commission will assess whether another proposal should be tabled or another type of approach should be chosen.”

For the privacy/cookie item (“concerning the respect for private life and the protection of personal data in electronic communications”), the EU said, “No foreseeable agreement – no agreement is expected from the colegislators. Furthermore, the proposal is outdated in view of some recent legislation in both the technological and the legislative landscape.”

Various EU specialists said those explanations were correct, but the reasons behind the decisions from those member countries were more complex. 

Andrew Gamino-Cheong, CTO at AI company Trustible, said different countries had different, and incompatible, positions.

“The EU member states have started to split on their own attitudes related to AI. On one extreme is France, which is trying to be pro-innovation and [French President Emmanuel] Macron used the [AI summit] this past week to emphasize that,” Gamino-Cheong said. “Others, including Germany, are very skeptical of AI still and were pushing for these regulations. If France and Germany are at odds, as the economic heavyweights in the EU, nothing will get done.”

But Gamino-Cheong, along with many others, said there is a fear that the global AI arms race may hurt countries that impose too many compliance requirements. 

The EU is seen as “being too aggressive, overregulating” and “the EU takes a 2-sentence description and writes 14.5 pages about it and then contradicts itself in multiple areas,” Gamino-Cheong said. 

Ian Tyler-Clarke, an executive counselor at the Info-Tech Research Group, said he was not happy that the two proposed bills did not go forward because he fears how those moves will influence other countries. 

“Beyond the EU, this decision could have broader geopolitical consequences. The EU has long been a global leader in setting regulatory precedents, particularly with GDPR, which influenced privacy laws worldwide. Without new AI liability rules, other regions may hesitate to introduce their own regulations, leading to a fragmented global approach,” Tyler-Clarke said. “Conversely, this could trigger a regulatory race to the bottom, where jurisdictions with the least restrictions attract AI development at the cost of oversight and accountability.”

A very different perspective comes from Enza Iannopollo, a Forrester principal analyst based in London. 

Asked about the failure to move forward on the privacy bill, Iannopollo said, “Thank God that they have withdrawn that one. There are more pressing priorities to address.”

She said the privacy effort suffered from the rapid advances in web controls, including some changes made by Google. “Regulators were not convinced that they would improve things,” Iannopollo said.

Regarding the AI liability rules, Iannopollo said that she expects to see those come back in a revised form. “I don’t think this is a final call. They are just buying time.”

The critical factor is that another, much larger piece of legislation, called simply the EU AI Act, is just about to kick in, and regulators wanted to see how that enforcement went before expanding it. “They want to see how these other pieces of the framework are going to work. There are a lot of moving parts so (delaying) is wise.”

Another analyst, Anshel Sag, VP and principal analyst with Moor Insights & Strategy, said that EU members are very concerned with how they are perceived globally.

“The real challenge is that applying regulations too early, without the industry being mature enough, risks hurting European companies and European competitiveness, which I believe is a major factor in why these regulations have been paused for now,” Sag said. “Especially when you consider the current rate of change within AI, there’s just a chance that they could spend a long time on this regulation and by the time it’s out, it’s already well out of date. They will have to act fast, though, when the time is right.”

Added Vincent Schmalbach, an independent AI engineer in Munich, “The most interesting part is how this represents a major shift in EU thinking. It went from being the world’s strictest tech regulator to acknowledging they need to focus on not falling further behind in the AI race.”

Michael Isbitski, principal application security architect for genAI at ADT, the $19 billion HR and payroll enterprise, and also a former Gartner analyst, sees the two proposed EU legislative efforts as potentially having had a massive impact on data strategies.

The proposed AI rule, he said, involved the retention of AI-generated data logs. “Everywhere there is some kind of AI transaction, you need to retain those logs, for every query, anywhere,” Isbitski said. “Think about what needs to be done to secure your requirements and controls systems, along with your cloud security. Logging seems simple, but if you look at a complete AI interaction, there are an awful lot of interconnects.”

However, Flavio Villanustre, global chief information security officer of LexisNexis Risk Solutions, said the pausing of these two EU potential rules will likely have no significant impact on enterprises.

“This means you can continue to do everything you were doing before. There will be no new constraints on top of anything you were doing,” Villanustre said.

But the broader issue of genAI liability absolutely needs to be addressed because the current mechanisms are woefully inadequate, he said. 

That is because the very nature of genAI, especially in its stochastic and probabilistic attributes, makes liability attribution virtually impossible.

Let’s say something bad happens, for example, with an LLM deployment where a company loses billions of dollars or there is a loss of life. 

There are typically going to be three possible groups to blame: the model-maker, which creates the algorithm and trains the model; the enterprise, which finetunes the models and adapts it to that enterprise’s needs; and the user base, which would be either employees, partners, or customers who pose the queries to the model.

Overwhelmingly, when a problem happens, it will be because of the interactions of efforts by two or three of those groups. Without the new legislation being proposed by the EU, the only means of determining liability will be via legal contracts. 

But genAI is a different kind of system. It can be asked the identical question five times and offer five different answers. That being the case, if its developers cannot accurately predict what it will do in different situations, Villanustre wondered what chance attorneys have at anticipating all problems.

“That is a challenge: determining who has the responsibility,” Villanustre said. “This legislation was meant to define the liability outside of contracts.”

Kategorie: Hacking & Security

It’s really happening: Are you prepared for the sunsetting of Exchange Server 2016 and 2019?

13 Únor, 2025 - 04:20

Enterprises still hosting on-premises Exchange mail, it’s time to face reality: Microsoft will soon no longer support your current licensing option.

Earlier this week, the tech giant released its final roll-up patch for Exchange Server 2019 (Cumulative Update 15). On October 14, Microsoft will officially stop supporting both Exchange Server 2016 and Exchange Server 2019 — meaning no more updates, technical help, bug fixes, or security patches.

With end of life (EOL) pending, Microsoft customers must now either move to the fully-hosted Exchange Online or Microsoft 365, or pay for on premises Exchange Server Subscription Edition (SE) licenses to receive continued support and updates.

“Every software company out there — Microsoft, Oracle — are all trying to nudge gently, and sometimes not so gently, their customer base to the cloud,” said Matt Kimball, a principal data center analyst with Moor Insights & Strategy. “They’re doing everything but forcing companies to move to cloud-based or SaaS-based models.”

Upgrade to Exchange Server SE, or stay out of date at your own risk

Microsoft will roll out Exchange Server SE early in the second half of 2025, and release the first CU for the platform later in the year. Once that happens, all other versions of Exchange Server will be out of support.

It’s not unexpected. Adam Preset, VP analyst at Gartner, pointed out that there have been 10 versions of Exchange Server since 1996. “This is just what happens,” he said.

The final cumulative update for Exchange 2019 integrates all prior security patches and introduces server-side components for Feature Flighting, an optional cloud-based service that allows for immediate updating once new features are available, he said. This will help ensure stability and security up to the EOL date.

“Post-EOL, organizations can operate existing installations at their own risk,” he said. “Email is an essential and business critical workload, though, so staying on Exchange Server 2019 is unwise.”

To install Exchange Server SE CU1 (or later), organizations will have to first decommission and remove all older versions of Exchange, according to Microsoft. Organizations have two options when moving to the new subscription  model: A legacy upgrade (introducing new servers and uninstalling old servers); or (only for 2019) an “in-place” upgrade (downloading and installing the latest upgrade package).

In addition to purchasing required licenses, customers must also maintain an active Microsoft subscription, which means purchasing either cloud subscription licenses for all users and devices, or buying Exchange Server SE licenses with software assurance (SA).

Preset pointed out that “there’s no substitute” for checking on licensing agreements and consulting with Microsoft if an organization needs to transition to Exchange Server SE. Also, the new model will accept Exchange 2019 product keys to help simplify the upgrade process.

It’s time for enterprises to embrace the future

To some, on-prem email hosting in 2025 seems like a quaint notion.

“Who the heck is still running Exchange on premises?,” Kimball asked. “I say that jokingly but I kind of mean it too.”

Cloud computing has grown massively over the last 15 years, and email has been typically one of the first candidates to be moved up into the cloud, he noted.

“Outside of super-high privacy reasons, I can’t see the efficiency of running on-premises,” said Kimball. However, he noted, “there’s always going to be a laggard customer base that is slow to adopt not even new technology, but current technology.”

These “corner cases” typically have unique privacy or regulatory requirements — or it might simply come down to company culture. “That points to something bigger: You’re hosting your own email and managing it a certain way because you’ve always done it that way,” said Kimball.

But that’s typically not best for a business’ users and partners, not to mention its IT staff, who want to be doing more exciting and challenging kinds of work. Enterprises should be focused on staying current, quickly gaining access to the latest capabilities (in Microsoft’s case, think access to Copilot) and the “absolute resiliency” of the cloud, said Kimball.

However, he did question Microsoft’s rather abrupt end-of-service for Exchange Server 2019. In other cases, even after EOL, Microsoft has been known to continue to support legacy infrastructures. “Five years is not a long time to have a product in the market,” said Kimball. “End of support is a big thing.”

Important migration strategies

Analysts emphasize that organizations on Exchange 2019 must build a strategy for migration, performing extensive planning and assessing infrastructure for a seamless transition.

The first thing to do, Kimball said: Perform a cost-benefit analysis or ROI study that takes into account all direct and indirect costs, infrastructure, software, people costs, financial impact of downtime — and what valuable IT staff have to do to maintain legacy environments “day after day after day, the care and feeding.” Then compare that to moving to hosted Exchange.

“Unless there’s a real hard and fast regulatory requirement, I would be willing to wager that the cost benefit analysis is going to lead to a migration to the cloud,” Kimball noted.

Overall, enterprises, whether Microsoft customers or not, should consider migrating to the cloud when on-premises maintenance costs rise, advanced security features are required, or they want more (and better) cloud integration, said Preset.

“Meticulous planning is essential to ensure a smooth transition,” he said.

To reduce the risks of disruption when migrating, Preset suggested having IT personnel skilled in Exchange and cloud technologies, project managers, and support from vendors with migration experience. Enterprises also need to allocate budget for new licenses, potential third-party migration tools, training expenses for IT staff, and new systems.

“However,” Preset emphasized, “if you know you need to transition to cloud anyway, Microsoft is not the only game in town. If you’re ready for a bigger change, it might be time to look at alternatives such as Google Workspace or other vendors with email services.”

Related:

Kategorie: Hacking & Security

AI company Ross Intelligence loses copyright fight with Thomson Reuters

12 Únor, 2025 - 17:59

A US judge has ruled in favor of Thomson Reuters in a AI training fight against Ross Intelligence, a legal AI startup, according to The Verge. Thomson Reuters sued Ross Intelligence in 2020 for using the company’s legal research platform Westlaw to train Ross Intelligence’s AI without permission. Westlaw indexes large amounts of non-copyrighted material, but mixes it with its own content.

Ross Intelligence argued that the training should be classified under “fair use” practices, but the judge disagreed. Instead, the court held that Ross Intelligence’s use of the copyrighted material affected its original value because the company intended to develop a direct competitor.

The ruling is significant because it could have implications for future cases where copyrighted material is used for AI training. One wrinkle: this particular case concerned non-generative AI, which is not the same as generative AI used in large language models to create new material based on previous training data.

Kategorie: Hacking & Security

BBC: Chatbots distort the facts about news

12 Únor, 2025 - 17:52

It’s already known that today’s generative AI (genAI) tools often have trouble with basic facts. Now, it’s clear they don’t well with current events either.

That’s the upshot of a test by the BBC, which asked ChatGPT, Copilot, Gemini and Perplexity to answer 100 questions using BBC articles as a source; more than half of the answers (51%) were wrong.

One in five answers (19%) were based on directly incorrect facts — and 13% of quotes had been modified from the source.cFor example, the AI tools believe that Rishi Sunak is still the UK’s Prime Minister, and they gave the wrong death date for TV personality Michael Mosley.

“The price of AI’s extraordinary benefits must not be a world where people searching for answers are served distorted, faulty content that appears to be fact,” Deborah Turness, managing director of BBC News, wrote. “In what can feel like a chaotic world, it really can’t be right that consumers seeking clarity are met with yet more confusion.”

Kategorie: Hacking & Security

Apple’s Chinese AI problem (perhaps) solved with Alibaba

12 Únor, 2025 - 17:42

Reflecting the erosion of universality, Apple Intelligence will now be coming to China, but rather than working with a US AI partner, the company will use Chinese-made AI tech from Alibaba.

According to The Information, Apple and Alibaba have already “submitted the co-developed features for approval to regulators.” The claim hasn’t yet been confirmed. If true, it would mean Alibaba’s Qwen model will replace OpenAI’s ChatGPT as the go-to third-party service integrated with Apple Intelligence for China.

Why does this matter?

Due to local regulations, Apple needs a Chinese partner to offer its AI services in China. It had been expected to work with Baidu, but chose not to do so. Deepseek was also considered, but as a small start-up there were concerns it would not scale to meet Apple-driven demand.

Achieving an AI deal in China is strategically important to Apple, as the lack of Apple Intelligence has quelled demand for its iPhones. Morgan Stanley analyst Erik Woodring told clients: “Our survey work shows that Chinese iPhone users are not only more interested in access to genAI technology than US or European iPhone owners, but over 50% of Chinese iPhone owners cited the staggered rollout of Apple Intelligence as having a moderate to significant impact on their decision not to upgrade to a new iPhone this cycle.”

Apple still maintains strong customer loyalty and relatively consistent switching rates, and government subsidies may help stimulate sales, Morgan Stanley said. 

The hope is that the introduction of Apple Intelligence (for China) will unleash pent-up demand for iPhone among Chinese consumers who might have delayed purchases pending this introduction.

So, when is Apple Intelligence coming to China?

News of the deal emerges just weeks before April, when Apple has already suggested it will introduce Simplified Chinese localization for Apple Intelligence. While both Apple and Alibaba must now gain regulatory approval for their plan, the signal within the smoke suggests a Q2 introduction of Chinese Apple Intelligence support. This could prove a shrewd move that might yet stimulate iPhone sales there — particularly with the introduction of the more affordable iPhone SE. 

Together, Apple could gain twin benefits — hardware sales and Apple Intelligence proliferation. Of course, the mass market AI that wins the platform wars will be the version people use, and Apple is likely optimistic that its huge and loyal user base will use its generative AI tools. 

Alibaba might be the first Chinese AI service supplier to gain integration on the iPhone, but perhaps not be the last. “Based on Apple’s broader rhetoric about providing users AI choice, we could also see a scenario where Apple creates an initial AI partnership with Alibaba that eventually expands to other local Chinese cloud players over time,” said Morgan Stanley in a client note.

Bugs in the lotion

One thing worth watching is the extent to which Apple is able to maintain privacy in China — will the company still offer private cloud servers for some tasks or offload all complex requests to its third-party partner? Will it limit the features it makes available to Chinese users? Or has it reached a deal with Chinese regulators in which the private services it does offer are seen to be relatively non-threatening?

Whatever the cut of the cloth, the seeming simplicity of Apple Intelligence masks numerous decisions around weft and weave, and not all of these will be evident on first glance.

You can follow me on social media! Join me on BlueSky,  LinkedInMastodon, and MeWe

Kategorie: Hacking & Security

DeepSeek — Latest news and insights

12 Únor, 2025 - 14:42

DeepSeek, founded in 2023 by Liang Wenfeng, a Chinese entrepreneur, engineer and former hedge fund manager, is generating a lot of buzz — and for good reason. Here are five things that make it stand out (as well as a listing of the latest news and analysis about DeepSeek).

5 things you need to know about DeepSeek
  1. More accessibility and efficiency: DeepSeek is designed to be less expensive to train and use than many competing large language models (LLMs).  Its architecture allows for high performance with fewer computational resources, which is designed to lead to faster response times and less energy consumption.
  2. Open-source availability and rapid development DeepSeek is under active development with new models and features being released regularly.  Models are often available for public download (on Hugging Face, for instance), which encourages collaboration and customization. 
  3. Advanced capabilities: reasoning and multimodal learning Models like DeepSeek-R1 are designed with a focus on advanced reasoning capabilities, aiming to go beyond simple text generation. DeepSeek is expanding into multimodal learning, handling diverse input types such as images, audio and text for a more comprehensive understanding.
  4. Limitations: Bias and context Like all LLMs, DeepSeek is susceptible to biases in its training data. Some biases may be intentional for content moderation purposes, which raises important ethical questions. While efficient, DeepSeek could have limitations in handling extremely long texts or complex conversations.
  5. Architecture and performance DeepSeek uses  a “mixture of experts” architecture, employing specialized submodels for different tasks, enhancing efficiency and potentially reducing training data needs. DeepSeek has demonstrated competitive performance, comparable to established models in certain tasks, especially mathematics and coding.
DeepSeek news and analysis 3 reasons Microsoft needn’t fear DeepSeek

February 12. 2025: Microsoft may have the most to lose from DeepSeek’s arrival. It’s invested billions of dollars in AI already, and has said this year alone it will invest another $80 billion. Given that DeepSeek said it built its newest chatbot so cheaply, is Microsoft throwing billions of dollars away? Can it compete with a company that can build genAI at such a low cost?

AI chatbot war breaks out with DeepSeek debut, and the winner is … you

February 5, 2025: DeepSeek’s chatbot hit Apple’s App Store and Google Play Store and downloads almost immediately exceeded those of OpenAI’s ChatGPT. In short order, the DeepSeek-R1 AI model changed the genAI market, erasing $600 billion in market capitalization — the largest intraday decline in history — before markets started to recover.

Hackers impersonate DeepSeek to distribute malware

February 4, 2025: To make things worse for DeepSeek, hackers were found flooding the Python Package Index (PyPI) repository with fake DeepSeek packages carrying malicious payloads. According to a discovery made by Positive Expert Security Center (PT ESC), a campaign was seen using this trick to dupe unsuspecting developers, ML engineers, and AI enthusiasts looking to integrate DeepSeek into their projects.

How would a potential ban on DeepSeek impact enterprises?

February 4, 2025: European regulators joined Microsoft, OpenAI, and the US government inefforts to determine if DeepSeek infringed on any copyrighted data from any US technology vendor. The investigations could potentially lead to a ban on DeepSeek in the US and EU, impacting millions of dollars that enterprises are already pouring into deploying DeepSeek AI models.

The DeepSeek lesson -— success without relying on Nvidia GPUs

Feb. 3, 2025: During the past two weeks, DeepSeek unraveled Silicon Valley’s comfortable narrative about generative AI (genAI) by introducing dramatically more efficient ways to scale large language models (LLMs). Without billions in venture capital to spend on Nvidia GPUs, DeepSeek had to be more resourceful and learned how to “activate only the most relevant portions of their mode

Nvidia unveils preview of DeepSeek-R1 NIM microservice

Jan. 31, 2025: Nvidia stock plummeted after Chinese AI developer DeepSeek unveiled its DeepSeek-R1 LLM. Last week, the chipmaker turned around and announced the DeepSeek-R1 model is available as a preview NIM on build.nvidia.com. Nvidia’s inference microservice is a set of containers and tools to help developers deploy and manage gen AI models across clouds, data centers, and workstations.

Italy blocks DeepSeek due to unclear data protection

Jan. 31, 2025: Italy’s data protection authority Garante has decided to block Chinese AI model DeepSeek in the country. The decision comes after the Chinese companies providing the chatbot service failed to provide the authority with sufficient information about how users’ personal data is used.

How DeepSeek changes the genAI equation for CIOs

Jan. 30, 2025: The new genAI model’s explosion on the scene is likely to amp up competition in the market, drive innovation, reduce costs and make gen AI initiatives more affordable. It’s also a metaphor for increasing disruption. Maybe it’s time for CIOs to reassess their AI strategies.

DeepSeek leaks 1 million sensitive records in a major data breach

Jan. 30, 2025: A New York-based cybersecurity firm, Wiz, has uncovered a critical security lapse at DeepSeek, a rising Chinese AI startup, revealing a cache of sensitive data openly accessible on the internet. According to Wiz, the exposed data included over a million lines of log entries, digital software keys, backend details, and user chat history from DeepSeek’s AI assistant.

Microsoft first raises doubts about DeepSeek and then adds it to its cloud

Jan. 30, 2025: Despite initiating a probe into the Chinese AI startup, Microsoft added DeepSeek’s latest reasoning model R1 to its model catalog on Azure AI Foundry and GitHub.

How DeepSeek will upend the AI industry — and open it to competition

Jan. 30, 2025: DeepSeek is more than China’s ChatGPT. It’s a major step forward for global AI by making model building cheaper, faster, and more accessible, according to Forrester Research. While LLMs aren’t the only route to advanced AI, DeepSeek should be “celebrated as a milestone for AI progress,” the research firm said.

DeepSeek triggers shock waves for AI giants, but the disruption won’t last

Jan. 28, 2025: DeepSeek’s open-source AI model’s impact lies in matching US models’ performance at a fraction of the cost by using compute and memory resources more efficiently. But industry analysts believe investor reaction to DeepSeek’s impact on US tech firms is being exaggerated.

DeepSeek hit by cyberattack and outage amid breakthrough success

Jan. 28, 2025: Chinese AI startup DeepSeek was hit by a cyberattack, according to the company, prompting it to restrict user registrations and manage website outages as demand for its AI assistant soared. According to the company’s status page, DeepSeek has been investigating the issue since late evening Beijing time on Monday.

What enterprises need to know about DeepSeek’s game-changing R1 AI model

Jan. 27, 2025: Two years ago, OpenAI’s ChatGPT launched a new wave of AI disruption that left the tech industry reassessing its future. Now, within the space of a week, a small Chinese startup called DeepSeek has pulled off a similar coup, this time at OpenAI’s expense.

iPhone users turn on to DeepSeek AI

Jan. 27, 2025: As if from nowhere, OpenAI competitor DeepSeek has risen to the top of the iPhone App Store chart, overtaking ChatGPT’s OpenAI. It’s the latest in a growing line of genAI services and seems to offer some significant advantages, not least its relatively lower development and production costs. 

Chinese AI startup DeepSeek unveils open-source model to rival OpenAI o1

Jan. 23, 2025: Chinese AI developer DeepSeek has unveiled an open-source version of its reasoning model, DeepSeek-R1, featuring 671 billion parameters and claiming performance superior to OpenAI’s o1 on key benchmark. “DeepSeek-R1 achieves a score of 79.8% Pass@1 on AIME 2024, slightly surpassing OpenAI-o1-1217,” the company said in a technical paper. “On MATH-500, it attains an impressive score of 97.3%, performing on par with OpenAI-o1-1217 and significantly outperforming other models.”

Kategorie: Hacking & Security

3 reasons Microsoft needn’t fear DeepSeek

12 Únor, 2025 - 12:00

The release of the latest version of the Chinese genAI bot DeepSeek last month upended the tech world when its creators claimed it was built for only $6 million — far less than the hundreds of billions of dollars Microsoft, Google, OpenAI, Meta, and others have poured into genAI development. 

The shockwaves were immediate. GenAI-related stocks took a nosedive, losing hundreds of billions of dollars in value overnight. Many prognosticators said DeepSeek would undermine America’s genAI dominance — and threaten the country’s big AI companies, notably Microsoft.

[ Related: More DeepSeek news and analysis ]

Microsoft, which became a $3-trillion company based on its AI leadership, has perhaps the most to lose from DeepSeek’s arrival. It’s invested billions of dollars in AI already, and has said this year alone it will invest another $80 billion. Given that DeepSeek said it built its newest chatbot so cheaply, is Microsoft throwing billions of dollars away? Can it compete with a company that can build genAI at such a low cost?

Microsoft has nothing to fear from DeepSeek. Here are three reasons the Chinese upstart won’t hurt Microsoft — and might even help it.

DeepSeek’s savings aren’t as large as it claims

DeepSeek’s claim that it developed the latest version of its chatbot for $6 million was eye-popping, given the amount of money being poured into AI development and related infrastructure by so many other companies. It was even more eye-popping because the chatbot appears to be technically on par with OpenAI’s ChatGPT, which underlies Microsoft’s Copilot.

But DeepSeek’s claim was extremely misleading. The semiconductor research and consulting firm  SemiAnalysis took a deep dive into the true costs of developing DeepSeek, based on information publicly provided by the Chinese company. SemiAnalysis found that the $6 million was “just the GPU cost of the pre-training run, which is only a portion of the total cost of the model. Excluded are important pieces of the puzzle like R&D and TCO of the hardware itself.”

Hardware costs, SemiAnalysis found, were likely well over a half billion dollars. It estimates that the total capital expenditure costs for the hardware, including the costs of operating it, were approximately $1.6 billion.

Beyond that, OpenAI claims that DeepSeek may have illegally used data created by OpenAI to train its model. The cost of obtaining training data can be billions of dollars, so we don’t know how much money DeepSeek would have had to spend if it didn’t use OpenAI’s data.

Although it’s still likely that DeepSeek spent much less than OpenAI, Microsoft, and competitors for building its model, its costs are likely in the billions of dollars, not a mere $6 million. And it’s not at all clear that DeepSeek can gain enough revenue to keep up its burn rate.

Businesses fear privacy and security breaches — and Chinese censorship

Cost savings are good. But even more important to most enterprises is the privacy and security of their data and business, and the privacy and security of their customers’ data.

Congress passed a law banning TikTok from the US based on fears that data is being gathered about users of the app and sent back to China. (US President Donald J. Trump has put a temporary hold on that ban.) But the kind of data that TikTok might gather and report back to China pales in comparison with the kinds of data DeepSeek might send. TikTok merely lets people post and watch videos. DeepSeek’s genAI chatbot has access to the sensitive personal, business, and financial data of enterprises and individuals that use it.

DeepSeek’s privacy policy admits upfront that it sends business and personal data to China, noting, “We store the information we collect in secure servers located in the People’s Republic of China.” The policy adds, “We may collect your text or audio input, prompt, uploaded files, feedback, chat history, or other content that you provide to our model and Services.”

Beyond that, Wired magazine adds: “DeepSeek says it will collect information about what device you are using, your operating system, IP address, and information such as crash reports. It can also record your ‘keystroke patterns or rhythms.’”

What might DeepSeek do with that data? Chinese companies are required by Chinese law to turn over any information to the Chinese government when requested. American businesses are unlikely to want to expose their data to the Chinese government in that way.

In addition, DeepSeek heavily censors its answers to requests, refusing to answer some questions, and providing Chinese propaganda for others, according to The New York Times. Businesses certainly don’t want to become arms of the Chinese government’s propaganda efforts.

Enterprises want off-the-shelf AI integration with business tools

What businesses want from genAI tools, above all, is to increase their productivity. Doing that requires integration with their applications, tools, and infrastructure. That’s exactly what Microsoft does with its entire Copilot product line, including Microsoft 365, OneDrive, SharePoint, Teams, GitHub, Microsoft’s CRM and ERP platform Dynamics 365, and others. 

DeepSeek offers nothing like that kind of integration. And without that, DeepSeek isn’t likely to make much progress against Microsoft — even if it can sell its chatbot more cheaply.

Microsoft itself doesn’t seem to be concerned, at least publicly. Microsoft CEO Satya Nadella even believes that the efficiencies DeepSeek has found in building AI will ultimately help his company’s bottom line.

“That type of optimization means AI will be much more ubiquitous,” he told Yahoo Finance. “And so, therefore, for a hyperscaler like us, a PC platform provider like us, this is all good news as far as I’m concerned.”

Kategorie: Hacking & Security

Yes, you can still upgrade Windows 10 PCs to Windows 11

12 Únor, 2025 - 12:00

Windows 10 has less than a year left before it hits its end of support deadline. Starting in October 2025, you’ll have to pay for security updates if you want to keep using Microsoft’s nearly-nine-year-old operating system. That means now is the time to think about upgrading any Windows 10 PCs you’re still working with to the current Windows 11 OS.

If you believe the viral headlines, things are getting messy: Microsoft, the rumors say, is actually trying to stop people from grabbing free upgrades to Windows 11, and the company is even eliminating a workaround that made that path possible. Could that really be true?

I’ll make it easy for you: That isn’t actually the case. You can absolutely still upgrade old and officially “unsupported” Windows 10 PCs to Windows 11, just as you could years ago when Windows 11 was released. Not much has changed.

So let’s look at what’s actually going on with Windows 11 upgrades in 2025. I’ll show you how you can still upgrade to Windows 11 — even if Windows Update says a system isn’t compatible and Microsoft doesn’t want to help. I’ll even explain why Windows 11 might not be the right fit for your PC.

That’s right: Even if you can, you might not want to upgrade after all — and that last part is what the controversy is really about.

Want to stay on top of what’s happening with Windows? Sign up for my free Windows Intelligence newsletter. I’ll send you free Windows Field Guide downloads as a special welcome bonus!

Windows 11 upgrade workarounds, explained

First things first: The newest Windows 10 PCs can easily upgrade to Windows 11 with no workarounds needed. If your PC is officially eligible for an easy upgrade, just open the Windows Update settings page on your Windows 10 PC. You’ll see a big message encouraging you to upgrade with a few clicks.

The oldest Windows 10 PCs, on the other hand, genuinely can’t upgrade to Windows 11 at all. They just don’t have the required hardware. Windows 11 needs Trusted Platform Module (TPM) hardware in order to operate, for one example, as it relies on that for certain hardware-based security functions. If your PC doesn’t have it, Windows 11 can’t run.

But there’s a mysterious third category of PCs in the middle. These PCs aren’t “officially” eligible for a supported upgrade, and Windows Update will never offer it. But they can run Windows 11. All you have to do is use a special registry hack while installing the software.

Consider the TPM hardware situation:

  • A PC without a TPM can’t upgrade to Windows 10.
  • A PC with TPM 2.0 hardware can upgrade to Windows 11 in the normal way.
  • But a PC with TPM 1.2 hardware? That PC can upgrade to Windows 11 — but only with the “AllowUpgradesWithUnsupportedTPMOrCPU” registry hack.

Microsoft has always warned that PCs upgraded using this registry hack workaround are technically “unsupported.” Microsoft says your PC may not work properly if you take that route and that it may one day stop offering Windows 11 updates to PCs that used the hack to upgrade. These warning messages date all the way back to the release of Windows 11. They’re nothing new.

Meanwhile, it’s worth noting that Microsoft is the one that made this registry hack workaround in the first place! It’s an “officially unofficial” way to get many Windows 10 PCs onto Windows 11 — without Microsoft’s guaranteed support and with a “your mileage may vary” warning — but with Microsoft’s help, in a roundabout way.

Microsoft’s hack-breaking mix-up

To be clear, Microsoft never encouraged average PC users to use the registry hack trick and upgrade their Windows 10 PCs to Windows 11. That path was intended more for Windows geeks and other technically inclined people. But, again, the company did create the registry hack, and it even provided instructions for following the procedure on its website — complete with warnings, naturally.

Even so, Microsoft doesn’t want to help people follow this path anymore. As spotted by Neowin earlier this month, Microsoft removed instructions for using the registry hack from its website. That’s it!

To be crystal clear, the registry hack still works. If you want to upgrade a Windows 10 PC to Windows 11, you can use the same registry hack you could’ve used two years ago. You’ll just have to find it from another source — not Microsoft.com.

Is it possible Microsoft might get rid of the workaround entirely? Sure. But there’s no indication that will happen. Instead, it just doesn’t want to encourage average PC users to try this tactic.

If you’re an average person looking to keep getting security updates for your Windows 10 PC after October, Microsoft would much prefer you buy a new Windows 11 PC — or pay $30 for another year of security updates.

There was also a recent story about Microsoft’s Defender antivirus blocking a tool that helps bypass these Windows 11 system requirements. For a few days, the “Flyby11” tool was flagged as malware. That’s changed: Defender doesn’t block this application anymore. And, even if it did, this tool is only one of many ways to upgrade an unsupported Windows 11 PC.

The reality of Microsoft’s Windows 10 upgrade warnings

In an update to the official support page in question, Microsoft explains its position:

“This support article was originally published on September 30, 2021, when Windows 11 was first released to the public. At the time of publication and still today, the intention behind this support page is to detail ways of installing Windows 11 on devices that meet system requirements for Windows 11. If you installed Windows 11 on a device not meeting Windows 11 system requirements, Microsoft recommends you roll back to Windows 10 immediately.

“Windows 11 minimum system requirements remain unchanged….”

See? There’s nothing really new here. Microsoft hasn’t changed anything about Windows; all it did was modify a web page. Yes, it recommends you roll back to Windows 10 if you’ve upgraded a PC with the registry hack. It’s always recommended you do so and, as the official guidance goes, avoid this registry hack.

Even so, countless people upgraded their old Windows 10 PCs to Windows 11 with this workaround. And I’ve yet to hear a single person who’s experienced a major issue after doing so.

If Microsoft were to change things in the future, the move would instantly break lots of existing Windows 11 PCs. That’d be a huge deal and the kind of controversy the company would likely rather avoid.

All this being said, you might want to at least think twice before rolling the dice on an unsupported upgrade. There’s a strong argument to be made for shielding yourself from even a potential mess, especially when it comes to a work-connected system. You could instead consider getting a new Windows 11 PC, sticking with Windows 10 and paying for security updates, or installing Linux or ChromeOS Flex to keep your PC running.

How to upgrade a Windows 10 PC to Windows 11

If, in spite of Microsoft’s warnings, you do want to upgrade an unsupported Windows 10 PC to Windows 11, the simplest way is to use the convenient Rufus tool to create a USB drive that’ll handle the installation and use the registry hack to skip the compatibility check at the same time.

Rufus offers a user-friendly way to use the Microsoft-created upgrade workaround.

Chris Hoffman, IDG

This won’t work with all Windows 10 PCs, but it will work with many of them — even if Windows Update tells you otherwise.

At the end of the day, remember: Microsoft may warn you that you’re on your own if you do this, but it’s always issued that warning. It’s up to you to decide which path you want to take, just as it has been since the start of this situation.

Let’s stay in touch! Sign up for my free Windows Intelligence newsletter. I’ll send you three new things to try each Friday and free Windows Field Guides as a special welcome gift.

Kategorie: Hacking & Security

The irritating but amusing irony of Google’s Gemini interface

12 Únor, 2025 - 11:45

Look, if you’ve read this column for long now, you know I’m extremely guarded with my enthusiasm for Gemini and the other similar large-language-model AI answer-bots.

Plain and simple, they just aren’t reliable as on-demand answer genies — despite being positioned as exactly that — and they’ve got a nasty habit of coughing up inaccurate info with an astonishing amount of confidence.

I’ve said it before, and I’ll say it again: If something is inaccurate or unreliable even 10% of the time (and that’s being generous, in this instance), it’s useful precisely 0% of the time.

But that foundational flaw isn’t what I want to talk about today — ’cause the truth is that for all of their weaknesses as information-surfacing systems, Gemini and its brethren do offer some genuine utility when it comes to other, more clerical functions. And plenty of folks are finding ways to work ’em into their workflow with lower-level tasks such as sorting through data and formatting spreadsheets (to name just a couple quick examples).

Clearly, Google wants Gemini to become an indispensable part of our lives both professionally and personally, as evidenced by the way it’s Google+’ing the service into our beaks at every possible opportunity. For as useful as it can be in certain limited scenarios, though, I can’t help but think Google is shooting itself in the foot with the way it’s presenting Gemini — in what’s an almost shockingly obvious-seeming miss, especially when it comes to the kind of more mainstream, not-just-early-adopter embracing the company is clearly aiming to achieve.

Let me show ya what I mean.

[Get level-headed knowledge in your inbox with my free Android Intelligence newsletter. Tips, insights, and other tasty treats await!]

Google’s Gemini interface puzzle

Right now, when you go to open up Gemini on Android on one of Google’s own Pixel devices, you’re greeted by a screen that looks a little somethin’ like this:

The Google Gemini full-screen interface, as seen on Android.

JR Raphael, IDG

See that little tool-tip at the top? “More models available” — “choose the one that best fits my needs,” you say? Okay, cool. I can get on board on with that. Let’s see what’s available.

A quick tap on that top part of the screen, and….

Google’s Gemini model list, from the app’s Android interface.

JR Raphael, IDG

What. The. Schmidt. Is. This.

To clarify, this is the standard Google Gemini experience — not any sort of beta or early access setup. This is what anyone who buys a new Pixel phone or Samsung Galaxy gadget gets when they activate what’s now their device’s default and prominently promoted assistant service. It’s also what anyone with any other Android device is now being pushed to use in place of the classic Google Assistant, with ever-increasing aggressiveness.

And I don’t think I can emphasize just how overwhelming of an interface is there and smacking you square in the peepers as soon as you take that step.

It’s not just on Android, either. You see the same selection when using Gemini on the web, too, maybe even with a more comically over-the-top appearance:

The Google Gemini interface in a desktop web browser.

JR Raphael, IDG

Seriously — what reasonably normal person who doesn’t work within Google’s engineering department could possibly parse this? And who would want to?

Making sense of the Gemini model mess

For context, what we’re seeing here is a list of every different version of Gemini Google’s released over recent months. Gemini 2.0 is the current version, launched just last week. Within that 2.0 framework, you’ve got four different possibilities to ponder:

  • “2.0 Flash” — “for everyday tasks, more features”
  • “2.0 Flash Thinking Experimental” — “best for multi-step reasoning”
  • “2.0 Flash Thinking Experimental with apps” — “reasoning across YouTube, Maps & Search”
  • And “2.0 Pro Experimental” — “best for complex tasks”

Erm, right.

Beyond that muddled mess, you can also choose to go back to the older “1.5 Pro with Deep Research” version of Gemini for “in-depth answers” as well as the “1.5 Pro” or “1.5 Flash” model. Sure — why not, right?

Let me be as blunt as I can be about this: Mushy-brained of a mammal as I may be, I’m someone who closely follows Google and studies its services as part of my job. I’m more tech-savvy than most average animals (which isn’t saying much, I realize, but even so). And I’ve been immersed in this particular part of the tech universe for something like 7,947 years now.

And yet, I couldn’t even begin to tell you what all that stuff means, in plain English, or why you might want to pick one Gemini model over another. Heck, even after reading Google’s 4,000-character oeuvre about all the ins and outs of this latest Gemini 2.0 edition, I couldn’t explain to you what, exactly, makes it any different to use on a practical level compared to the earlier versions — nor, after spending quite a bit of time testing it, could I identify to you how it’s done things any better for me in any meaningful, measurable, and specific way.

And that, m’dear, is absolutely hilarious to me — because the entire point of Gemini is that it’s supposed to help us understand stuff and make our lives easier. But somehow, its very interface is so frickin’ complex and convoluted that we practically need another version of Gemini just to decipher it and help us understand which version of Gemini we’re supposed to use for what and why.

The irony is delightful. But all bemused chuckling aside, it’s also a pretty serious problem.

I mean, really: Imagine one of your less savvy co-workers — or maybe even your wacky cousin Winslow from West Virginia — following Google’s prompts to try out Gemini and then encountering this monstrous menu of mumbo-jumbo. There’s no way they’d be able to make heads or tails of it, and I’d be willing to wager they’d just close the thing once and for all the second they saw it.

And that’s to say nothing of what happens if they actually make it past that first impression and then realize they’ve gotta keep wading through that labyrinth and figuring out the appropriate Gemini version every single time they come up with a new task or question.

It just isn’t a good experience by any measure — but especially not for a service that promises to save you time and simplify your life. And that, suffice it to say, doesn’t exactly jibe with Google’s goal of getting everyone in the habit of using Gemini constantly across all of their devices.

Here’s the bottom line: Gemini isn’t a beta-level experimental feature anymore. It’s a prominent public service — perhaps even Google’s most prominent product at the moment. It’s now a core part of the company’s enterprise-focused Workspace offering. For Goog’s sake, there was even a Super Bowl ad about it. For individuals and companies alike, it’s clearly meant to be serious business — and yet, it still feels like a clunky developer play-space.

If Google really wants people to accept Gemini as an everyday tool for workplace productivity and beyond, they’ve gotta make it more accessible and aimed at actual regular-human use. It needs to be intuitive, approachable, and easy for anyone to understand. And you don’t need an accuracy-challenged AI assistant to tell you this isn’t the way to achieve that.

Get practical tips, personal recommendations, and plain-English perspective on all the latest Googley twists with my free Android Intelligence newsletter — three new things to know and try each Friday.

Kategorie: Hacking & Security

Microsoft 365: A guide to the updates

12 Únor, 2025 - 11:18

Microsoft 365 (and Office 365) subscribers get more frequent software updates than those who have purchased Office without a subscription, which means subscribers have access to the latest features, security patches, and bug fixes. But it can be hard to keep track of the changes in each update and know when they’re available. We’re doing this for you, so you don’t have to.

Following are summaries of the updates to Microsoft 365/Office 365 for Windows over the past year, with the latest releases shown first. We’ll add info about new updates as they’re rolled out.

Note: This story covers updates released to the Current Channel for Microsoft 365/Office 365 subscriptions. If you’re a member of Microsoft’s Office Insider preview program or want to get a sneak peek at upcoming features, see the Microsoft 365 Insider blog.

Version 2501 (Build 18429.20158)

Release date: February 11, 2025

This build removes the option to display Track Changes balloons in left margin in Word. It also includes a variety of security updates. See “Release notes for Microsoft Office security updates” for details.

What IT needs to know: Because this is a security update, it should be applied relatively soon. Over the next few weeks, check for reports about problematic issues, and if all seems well, apply the update.

Get more info about Version 2501 (Build 18429.20158).

Version 2501 (Build 18429.20132)

Release date: January 30, 2025

In this build, the advanced Track Changes option to set the margin for balloons in Word has been removed.

A wide variety of bugs have also been fixed, including one in which ActiveX controls used an excessive amount of GDI handles in PowerPoint, and another for the entire Office suite in which images couldn’t be pasted from SharePoint.

 Get more info about Version 2501 (Build 18429.20132).

Version 2412 (Build 18324.20194)

Release date: January 16, 2025

This build fixes one bug, in which apps would exit unexpectedly when running on Windows Server 2016.

Get more info about Version 2412 (Build 18324.20194).

Version 2412 (Build 18324.20190)

Release date: January 14, 2025

This build fixes a bug in Word in which the layout of tables were changed unexpectedly. It also includes a variety of security updates. See Release notes for Microsoft Office security updates for details.

What IT needs to know: Because this is a security update, it should be applied relatively soon. Over the next few weeks, check for reports about problematic issues, and if all seems well, apply the update.

Get more info about Version 2412 (Build 18324.20190).

Version 2412 (Build 18324.20168)

Release date: January 7, 2025

This build makes tables in Outlook more accessible for screen readers. It also fixes a wide variety of bugs, including one in Word in which a document saved to a network shared folder and set to “Always Open Read-Only” would open in “Editing” mode, and another for the entire Office suite in which application didn’t render the grid properly after switching from page break preview to normal view.

Get more info about Version 2412 (Build 18324.20168).

Version 2411 (Build 18227.20162)

Release date: December 10, 2024

This build fixes a bug in Word and Outlook where characters didn’t render correctly when using Save Selection to Text Box Gallery. It also includes a variety of security updates. See Release notes for Microsoft Office security updates for details.

What IT needs to know: Because this is a security update, it should be applied relatively soon. Over the next few weeks, check for reports about problematic issues, and if all seems well, apply the update.

Get more info about Version 2411 (Build 18227.20162).

Version 2411 (Build 18227.20152)

Release date: December 5, 2024

This build fixes a wide variety of bugs, including one in Excel in which some cells might not be rendered properly upon scrolling in a worksheet using freeze panes, one in Word which prevented emails with linked SVG content from saving or sending, and one in which some PowerPoint presentations created by third-party tools didn’t open correctly and some content was removed.

Get more info about Version 2411 (Build 18227.20152).

Version 2410 (Build 18129.20158)

Release date: November 12, 2024

This build fixes a variety of bugs, including one in Word in which all characters didn’t appear correctly when creating an Outlook task from OneNote, and one in PowerPoint in which embedded BMP images in the PowerPoint slide were not opening.

This build also includes a variety of security updates. See Release notes for Microsoft Office security updates for details.

What IT needs to know: Because this is a security update, it should be applied relatively soon. Over the next few weeks, check for reports about problematic issues, and if all seems well, apply the update.

Get more info about Version 2410 (Build 18129.20158).

Version 2410 (Build 18129.20116)

Release date: October 28, 2024

This build enables filtering capabilities for the comment pane in Excel and fixes a variety of bugs, including one in Word in which the title bar no longer showed a “Saved” status for locally saved files, and one in PowerPoint in which a graphics-related issue caused the app to close unexpectedly at times.

Get more info about Version 2410 (Build 18129.20116).

Version 2409 (Build 18025.20160)

Release date: October 15, 2024

This build fixes a single bug in Word, in which emails with linked SVG content couldn’t be saved or sent.

Get more info about Version 2409 (Build 18025.20160).

Version 2409 (Build 18025.20140)

Release date: October 8, 2024

This build fixes a variety of bugs, including one in Word in which text wasn’t clearly visible in High Contrast Mode when using “Draft with Copilot” and referencing a meeting under “Reference your content.”

This build also includes multiple security updates. See Release notes for Microsoft Office security updates for details.

What IT needs to know: Because this is a security update, it should be applied relatively soon. Over the next few weeks, check for reports about problematic issues, and if all seems well, apply the update.

Get more info about Version 2409 (Build 18025.20140).

Version 2409 (Build 18025.20104)

Release date: September 25, 2024

This build fixes a single bug, in which when you saved a file in Word, the save status was missing from the Title bar.

Get more info about Version 2409 (Build 18025.20104).

Version 2409 (Build 18025.20096)

Release date: September 23, 2024

This build improves the user experience for selecting which users should have which permissions when a sensitivity label configured for user-defined permissions is applied to a file or when configuring standalone Information Rights Management through the Restrict Access feature. This change affects Excel, PowerPoint, and Word.

The build also fixes a variety of bugs, including one in Word in which Document Mode would switch from “editing” to “viewing” if user enabled “Track Changes” and set “For Everyone.”

Get more info about Version 2409 (Build 18025.20096).

Version 2408 (Build 17928.20156)

Release date: September 10, 2024

This update will remove Flip video support when the service goes offline on October 1, 2024. The build also includes a variety of security updates. Go here for details.

What IT needs to know: Because this is a security update, it should be applied relatively soon. Over the next few weeks, check for reports about problematic issues, and if all seems well, apply the update.

Get more info about Version 2408 (Build 17928.20156).

Version 2408 (Build 17928.20114)

Release date: August 26, 2024

This build allows you to disable connected experiences for privacy concerns without impacting data security policies, such as sensitivity labels. Services associated with Microsoft Purview (e.g., sensitivity labels and rights management) are no longer controlled by policy settings to manage privacy controls for Microsoft 365 Apps. Instead, these services will rely on their existing security admin controls in Purview portals.

The build also fixes a variety of bugs, including one in Outlook that caused default SMIME labels to fail to apply when a user replied to or forwarded an unlabeled message, and one for the entire suite in which people couldn’t install Microsoft 365 apps on an enrolled device.

Get more info about Version 2408 (Build 17928.20114).

Version 2407 (Build 17830.20166)

Release date: August 13, 2024

This build includes a variety of security updates for Excel, Outlook, PowerPoint, Project, Visio, and the entire Office suite. See Microsoft’s Release notes for Office security updates for details.

What IT needs to know: Because this is a security update, it should be applied relatively soon. Over the next few weeks, check for reports about problematic issues, and if all seems well, apply the update.

Get more info about Version 2407 (Build 17830.20166).

Version 2407 (Build 17830.20138)

Release date: August 1, 2024

This build fixes a wide variety of bugs, including one in which coauthoring on text boxes in Excel sometimes gave unexpected results, another in PowerPoint in which line widths were not preserved when exporting arrow shapes to PDF, and another in Word in which revisions were sometimes skipped when reviewing using VBA.

Get more info about Version 2407 (Build 17830.20138).

Version 2406 (Build 17726.20160)

Release date: July 9, 2024

This build fixes several bugs, including one in Word and Excel in which characters don’t appear correctly in Text Box Gallery. It also fixes a number of security holes. For details, see Release notes for Microsoft Office security updates.

What IT needs to know: Because this is a security update, it should be applied relatively soon. Over the next few weeks, check for reports about problematic issues, and if all seems well, apply the update.

Get more info about Version 2406 (Build 17726.20160).

Version 2406 (Build 17726.20126)

Release date: June 26, 2024

This build fixes a wide variety of bugs, including one in which Excel documents might be unexpectedly edited when a mandatory sensitivity label has not been applied, one that caused Outlook to exit unexpectedly shortly after launch for some users, and one in which pasting data from Word or Excel to an Outlook template as a link would cause an error message to appear.

Get more info about Version 2406 (Build 17726.20126).

Version 2405 (Build 17628.20164)

Release date: June 19, 2024

This build includes a variety of unspecified bug and performance fixes.

Get more info about Version 2405 (Build 17628.20164).

Version 2405 (Build 17628.20144)

Release date: June 11, 2024

This build fixes one bug, which prevented users from sending mail for a few hours after updating add-ins with on-send events. It also fixes a number of security holes. For details, see Release notes for Microsoft Office security updates.

What IT needs to know: Because this is a security update, it should be applied relatively soon. Over the next few weeks, check for reports about problematic issues, and if all seems well, apply the update.

Get more info about Version 2405 (Build 17628.20144).

Version 2405 (Build 17628.20110)

Release date: May 30, 2024

This build fixes a wide variety of bugs, including one in Excel in which an embedded workbook in .xls format might not have closed properly, one that that caused Outlook to close when using Copilot Summarize, one in Word in which content controls may have been removed when coauthoring, and one for the entire Office suite in which the Organization Chart Add-In for Microsoft programs was not loading properly.

Get more info about Version 2405 (Build 17628.20110).

Version 2404 (Build 17531.20152)

Release date: May 14, 2024

This build fixes a number of bugs, including one in Word where content controls might be removed when coauthoring, and one that caused Sovereign users to be unable to create ToDo tasks from Outlook.

It also fixes a number of security holes. For details, see Release notes for Microsoft Office security updates.

What IT needs to know: Because this is a security update, it should be applied relatively soon. Over the next few weeks, check for reports about problematic issues, and if all seems well, apply the update.

Get more info about Version 2404 (Build 17531.20152).

Version 2404 (Build 17531.20140)

Release date: May 7, 2024

This build fixes two bugs in Outlook, one in which it closed unexpectedly using the Scheduling Assistant when creating a new meeting or viewing an existing meeting, and another that caused add-in developers to hit timeouts when retrieving notifications from an Outlook client context.

Get more info about Version 2404 (Build 17531.20140) .

Version 2404 (Build 17531.20120)

Release date: April 29, 2024

This build reduces workbook size bloat from unnecessary cell formatting with a new “Check Performance” task pane. In addition, it fixes a wide variety of bugs, including one in Excel in which the default font could not be set; one in Outlook in which custom forms from MAPI form servers stopped responding; one in PowerPoint in which online videos did not play in some cases; one in which when opening certain Word documents would cause the error, “Word experienced an error trying to open the file”; and one in which the Office update installer appeared to be unresponsive.

Get more info about Version 2404 (Build 17531.20120) .

Version 2403 (Build 17425.20176)

Release date: April 9, 2024

This build fixes a number of security holes. For details, see Release notes for Microsoft Office security updates.

What IT needs to know: Because this is a security update, it should be applied relatively soon. Over the next few weeks, check for reports about problematic issues, and if all seems well, apply the update.

Get more info about Version 2403 (Build 17425.20176).

Version 2402 (Build 17328.20184)

Release date: March 12, 2024

This build fixes three bugs: one in which Access closed unexpectedly, one in which Excel closed unexpectedly when opening files with pivot tables and table design in macro-enabled files, and one in which Word closed unexpectedly when the undo function was used.

This build also fixes a number of security holes. For details, see Release notes for Microsoft Office security updates.

What IT needs to know: Because this is a security update, it should be applied relatively soon. Over the next few weeks, check for reports about problematic issues, and if all seems well, apply the update.

Get more info about Version 2402 (Build 17328.20184).

Version 2402 (Build 17328.20162)

Release date: March 4, 2024

This build fixes several bugs, including one that crashed Outlook when a link was clicked on, and another for the entire Office suite in which opened Office apps didn’t automatically start when a laptop was reopened, and an error message appeared after manual relaunch.

Get more info about Version 2402 (Build 17328.20162).

Version 2402 (Build 17328.20142)

Release date: February 28, 2024

This build fixes a variety of bugs, including one that caused Outlook to exit unexpectedly when expanding a conversation in the search results from a search of “All Mailboxes,” and another in which users were not able to create a bullet list with hyphens in PowerPoint.

Get more info about Version 2402 (Build 17328.20142).

Version 2401 (Build 17231.20236)

Release date: February 13, 2024

This build fixes several bugs, including one in which macros were being corrupted when saving Excel files and another that affected the entire Office suite in which add-ins would not load after Click trust for content add-in was selected.

This build also fixes a number of security holes. For details, see Release notes for Microsoft Office security updates.

What IT needs to know: Because this is a security update, it should be applied relatively soon. Over the next few weeks, check for reports about problematic issues, and if all seems well, apply the update.

Get more info about Version 2401 (Build 17231.20236).

Version 2401 (Build 17231.20194)

Release date: February 1, 2024

This build fixes a single bug in which expanded groups in the message list collapsed when users changed which column they were arranged by.

Get more info about Version 2401 (Build 17231.20194).

Version 2401 (Build 17231.20182)

Release date: January 30, 2024

This build fixes a wide variety of bugs, including one in which Excel would stop responding when saving changes, one in PowerPoint in which Notes and Slide layout would open with incorrect proportions when a file was opened from a protected view, and one in Word in which comment cards appeared too wide and cut off text when changing or switching the screen in use.

Get more info about Version 2401 (Build 17231.20182).

Version 2312 (Build 17126.20132)

Release date: January 9, 2024

This build fixes a number of security holes. For details, see Release notes for Microsoft Office security updates.

What IT needs to know: Because this is a security update, it should be applied relatively soon. Over the next few weeks, check for reports about problematic issues, and if all seems well, apply the update.

Get more info about Version 2312 (Build 17126.20132).

Version 2312 (Build 17126.20126)

Release date: January 4, 2023

This build introduces a new sensitivity toolbar in Word, Excel, and PowerPoint that helps users understand the security policies that apply to their documents. It’s available when users are creating copies of their documents in File / Save As. In addition, Office now had a new default theme, which Microsoft says is “more modern and accessible.”

It also fixes a wide variety of bugs, including one in Excel in which Custom Menu text was truncated when right-clicking in a cell, one in PowerPoint in which restoring a previous version of a presentation was not working as expected when using Version History, and one in Word in which the content control end tag was marked at the end of the document automatically if the document was edited in Word Online and then opened in Word desktop.

Get more info about  Version 2312 (Build 17126.20126).

Version 2311 (Build 17029.20108)

Release date: December 12, 2023

This build fixes one bug in Outlook, in which the message list was blank when switching between the “Focused” and “Other” views.

It also fixes a number of security holes. For details, see Release notes for Microsoft Office security updates.

What IT needs to know: Because this is a security update, it should be applied relatively soon. Over the next few weeks, check for reports about problematic issues, and if all seems well, apply the update.

Get more info about Version 2311 (Build 17029.20108).

Version 2311 (Build 17029.20068)

Release date: November 29, 2023

This build automatically inserts image captioning for Excel’s images. When you insert an image into a spreadsheet, accessibility image captioning is automatically generated for you.

It also fixes a wide variety of bugs, including one in Excel in which list box controls would not respond to mouse clicks after scrolling using the mouse wheel, and one in Word in which the language of a presentation was not retained when saving or exporting the presentation to a PDF file.

Get more info about Version 2311 (Build 17029.20068).

Version 2310 (Build 16924.20150)

Release date: November 14, 2023

This build fixes several bugs, including one in which Outlook failed to comply with the default browser settings for some users, and another in which new lines were added to an Outlook signature when pressing Enter in the body of the email.

It also fixes a number of security holes. For details, see Release notes for Microsoft Office security updates.

What IT needs to know: Because this is a security update, it should be applied relatively soon. Over the next few weeks, check for reports about problematic issues, and if all seems well, apply the update.

Get more info about Version 2310 (Build 16924.20150).

Version 2310 (Build 16924.20124)

Release date: Oct. 31, 2023

This build fixes a bug that caused Outlook to exit unexpectedly when clicking the More link in the Search results list.

Get more info about Version 2310 (Build 16924.20124).

Version 2310 (Build 16924.20106)

Release date: Oct. 25, 2023

In this build, the Teams Meeting App works in Outlook, too. With it, you’ll be able to configure a meeting app while scheduling an invite in Outlook. The meeting app will be ready to use when you chat or join the meeting on Teams.

A wide variety of bugs have also been fixed, including one in Excel where certain Pivot Tables would load slowly; one in which OneNote would close unexpectedly when rapidly navigating from one .PDF file to another .PDF file between different sections, or when performing an undo operation on a .PDF printout insertion; and one in the entire Office suite that caused unexpected black borders to appear around screen captures added with the Insert Screenshot functionality.

Get more info about Version 2310 (Build 16924.20106).

Version 2309 (Build 16827.20166)

Release date: October 10, 2023

This build fixes two bugs, one in which users were missing their Outlook add-ins, and another in Word in which subheading numbering with a custom Style would disappear if the file was saved and reopened. It also fixes a number of security holes. For details, see Release notes for Microsoft Office security updates.

What IT needs to know: Because this is a security update, it should be applied relatively soon. Over the next few weeks, check for reports about problematic issues, and if all seems well, apply the update.

Get more info about Version 2309 (Build 16827.20166).

Version 2309 (Build 16827.20130)

Release date: September 28, 2023

This build introduces two new features, including the ability to disable specific types of automatic data conversions in Excel and support for the “Present in Teams” button to present local files in PowerPoint Live in Microsoft Teams.

Several bugs have also been fixed, including one in which the setting to control how Outlook opens previous items at start-up was missing from the Options window, and another in Word in which the Add-ins tab was not visible when using custom toolbar information.

Get more info about Version 2309 (Build 16827.20130).

Version 2308 (Build 16731.20234)

Release date: September 12, 2023

This build fixes several bugs, including one that caused Outlook to close unexpectedly when viewing an email, and another in PowerPoint in which the presenter view slide section zoomed in and out when zooming in the notes section.

It also fixes a number of security holes. For details, see Release notes for Microsoft Office security updates.

What IT needs to know: Because this is a security update, it should be applied relatively soon. Over the next few weeks, check for reports about problematic issues, and if all seems well, apply the update.

Get more info about Version 2308 (Build 16731.20234).

Kategorie: Hacking & Security

Enterprise tech spending to hit $4.9 trillion in 2025, driven by AI, cloud, and cybersecurity

12 Únor, 2025 - 11:15

Global enterprise technology spending is set to grow by 5.6% in 2025, reaching $4.9 trillion, as enterprises continue to prioritize investments in cybersecurity, cloud computing, generative AI, and digital transformation.

North America and the Asia-Pacific region are projected to be the fastest-growing markets, while software and IT services are projected to account for 70% of all global technology spending by 2029, according to a Forrester report.

“Despite geopolitical instability and a softening IT and telecom services market in 2024, technology investments remain resilient,” said the report titled Global Tech Market Forecast, 2024-2029.

While certain sectors of the IT and telecom services market are showing signs of slowing down, businesses are accelerating their adoption of AI-driven tools and cloud-based solutions to enhance productivity and efficiency.

“Over the next five years, technology investments will reshape industries at an unprecedented pace,” Michael O’Grady, principal forecast analyst at Forrester said in the report. “GenAI, cloud technologies, and cybersecurity will take center stage, transforming how businesses operate and deliver value.”

O’Grady further said that companies that prioritize these investments “will not only strengthen their competitive edge but also achieve sustainable growth, but it’s important that they also balance their rapid tech investments with ongoing efforts to manage legacy systems and reduce technical debt.”

Software and AI investments fuel growth

Forrester projects that software spending will grow by 10.5% in 2025, making it the fastest-growing category within the global tech market. Enterprise investments in AI, cloud computing, and cybersecurity are expected to drive long-term expansion, with businesses increasingly shifting toward SaaS-based models.

Software will comprise 37% of global technology spending by 2029, nearly doubling its share from 2016, the report noted.

“The balance between AI hype and enterprise adoption is stabilizing as businesses focus on practical, ROI-driven applications,” said Charlie Dai, VP and principal analyst at Forrester.

He noted that while AI spending continues to grow, its measurable benefits are now evident in areas such as document automation, customer service, and employee augmentation. “Success depends on clear use cases, integration, and managing expectations, ensuring investments align with tangible business outcomes,” Dai said.

The report further added that with the demand for AI-driven infrastructure rising rapidly, AI server and storage markets are expected to see a 13% annual growth rate through 2030. OpenAI’s annualized revenue has already surged to $3.4 billion, up from $1 billion in mid-2023, highlighting the increasing adoption of generative AI solutions in enterprise environments.

This trend reflects broader corporate interest in AI-powered automation, which is transforming industries ranging from healthcare and finance to manufacturing and retail.

Cloud transformation on the rise

The IT services sector is expected to grow by 3.6% in 2025, as businesses continue to rely on consulting, IT outsourcing, and infrastructure-as-a-service (IaaS) to modernize their operations. The shift from traditional capital expenditures to operating expenditures through cloud-based services is accelerating, as enterprises seek more scalable and cost-effective solutions.

Dai noted that while many industries are shifting to an opex model for flexibility and scalability, a hybrid approach will persist. He explained that sectors such as manufacturing and utilities will likely continue investing in capex for critical, long-term infrastructure, whereas tech-driven industries will favor opex-based cloud solutions.

“Cost control, regulatory requirements, and strategic asset ownership will drive this decision,” he said.

IaaS is poised for substantial growth, with a projected compound annual growth rate of 16% through 2028. This expansion is being driven by enterprises migrating workloads to major cloud providers, including Microsoft Azure, AWS, and Google Cloud, the report added.

These investments are expected to improve operational agility, reduce infrastructure costs, and enhance security resilience in an increasingly digital business landscape.

Europe faces challenges

According to the report, technology spending trends vary significantly by region, with North America and Asia-Pacific expected to see the strongest growth.

North America is projected to experience a 6.1% increase in tech spending, with AI investments in financial services, retail, and media leading the way. Businesses in the US and Canada are accelerating cloud migration and cybersecurity initiatives, positioning the region at the forefront of enterprise IT innovation.

The Asia-Pacific region is forecasted to grow by 5.6%, with China, India, Japan, and Malaysia emerging as key drivers of expansion. India, in particular, is expected to have the region’s fastest-growing tech spending CAGR of 9.6% from 2024 to 2029, fueled by investments in AI, cloud, and digital transformation initiatives. The region’s investments in AI and semiconductor technologies continue to support enterprise adoption of next-generation computing solutions.

Meanwhile, Europe’s tech market is expected to grow at a slower rate of 5%, as economic challenges in Germany and Italy dampen enterprise spending.

Charlie Dai pointed to fragmented regulations, stricter data privacy laws, and higher operational costs as key barriers to faster enterprise IT growth in Europe.

He explained that cultural diversity and varying levels of digital maturity across countries further complicate scaling efforts for technology providers. “Europe faces slower enterprise IT growth due to fragmented regulations, stricter data privacy laws, and higher operational costs,” he said.

Latin America and the Middle East are also witnessing steady growth, with tech spending in these regions projected to rise between 5.2% and 5.4%. Governments and telecom operators are leading digital transformation efforts, with cloud adoption and AI integration playing a crucial role in modernizing public services and business operations.

Enterprise takeaways and strategic considerations

For enterprises, the forecast highlights several key takeaways that will shape IT investment strategies in the coming years. The growing dominance of cloud and AI-driven technologies is compelling organizations to rethink their approach to IT spending. Cybersecurity remains a top priority, with leading firms such as Palo Alto Networks forecasting a 16% revenue increase in 2024, reflecting heightened enterprise demand for advanced security solutions.

“Enterprises should prioritize a zero-trust strategy, integrating cybersecurity and compliance into every stage of AI and cloud adoption,” Dai said, adding that “enterprises must rethink their IT investment strategies by integrating cybersecurity and compliance into every stage of AI and cloud adoption.”

At the same time, organizations must navigate the challenges of modernizing legacy IT infrastructure. While cloud adoption is accelerating, two-thirds of global IT budgets are still allocated to maintaining existing systems. This underscores the complexity of balancing innovation with managing costs, ensuring compliance, and mitigating security risks.

The evolving regulatory landscape around AI and data protection is further influencing how enterprises deploy new technologies while safeguarding customer trust and operational integrity. With enterprises expected to spend nearly $5 trillion on technology in 2025, the decisions made today will have a lasting impact on business resilience and digital competitiveness.

Kategorie: Hacking & Security

Will the non-English genAI problem lead to data transparency and lower costs?

12 Únor, 2025 - 11:00

It’s become increasing clear that quality plunges when moving from English to non-English-based large language models (LLMs). They’re less accurate and there’s a serious lack of transparency around data training, both in terms of data volume and data quality.

The latter has long been a problem for generative AI (genAI) tools and platforms.

But enterprises aren’t paying less for less-productive models, even though the value they offer is diminished. So, why aren’t CIOs getting a price break for non-English models? Because without any data transparency, they rarely know they’re paying more for less. 

There are a variety of reasons why model makers don’t disclose their data training particulars. (Let’s not even get into the issue of whether they have legal rights to do whatever training they did — though it’s tempting to do so, if only to explore the hypocrisy of OpenAI complaining about DeepSeek not getting permission before training on much of its data.) 

Speaking of DeepSeek, don’t read too much into the lower cost of its underlying models. Yes, its builders cleverly leveraged open source to find efficiencies and lower pricing, but there’s been little disclosure of how much the Chinese government helped with DeepSeek’s funding, either directly or indirectly. 

That said, if DeepSeek is the cudgel that puts downward pressure on genAI pricing, I’m all for it — and IT execs should be, too. But until we see evidence of meaningful price cuts, they should use the lack of data transparency in non-English models to try and get model maker pricetags out of the stratospheric. 

The non-English issue isn’t really about the language, per se. It’s more about the training data that is available within that language. (By some estimates, the training datasets for non-English models could be just 1/10 or even 1/100 the size of their English counterparts.)

Hans Florian, whose title is a distinguished research scientist for multilingual natural language processing at IBM, said he uses a trick to guesstimate how much data is available in various languages. “You can look at the number of Wikipedia pages in that language. That correlates quite well with the amount of data available in that language,” he said.

To further complicate the issue, sometimes it’s not about the language or the available data in that language. It can — logically enough — be about data related to activities in the region where a particular  language is dominant.

If model makers start seeing meaningful pricing pushback from a lot of enterprises concerned about model quality, they have only a couple of options. They can selectively — and secretly — negotiate lower prices for non-English models for some of their customers — or they can get serious about data transparency.

Because LLM makers have invested billions of dollars in genAI, they aren’t going to like the idea of lower pricing. That leads to that second option: deliver full transparency to all customers about all models — both in terms of quantity and quality — and price their wares accordingly. 

Given that quality is almost impossible to represent numerically, that will mean disclosing all training data details so each customer can make their own determination of quality for the topics, verticals and geographies they care about.

The pricing disparity between what a model can deliver and what an enterprise is forced to pay is at the heart of why CIOs are still struggling to deliver genAI ROI

Obviously, lower pricing would be the best way to improve the ROI for genAI investments. But if that’s not going to happen anytime soon, full data transparency is the next best thing.

There is a catch: model makers almost certainly realize that full data-training transparency will likely force them to lower prices, since it would showcase how low quality their data is. 

Note: I say that their data is low-quality as if it’s a given; it is absolutely a given. If model makers believed they were using lots of high-quality data, far from resisting transparency, they would embrace it. It would be a selling point. It might even be useful for propping up prices. High quality usually sells itself.

Their refusal to deliver any kind of data-training transparency tells you everything you need to know about their quality beliefs, and about the state of the market at the moment. 

Kategorie: Hacking & Security

Monday.com aims to be an ‘AI-First’ platform with latest enhancements

12 Únor, 2025 - 03:57

Monday.com has accelerated its push into artificial intelligence (AI) with the announcement this week of its AI vision, which includes three areas of focus: AI Blocks, embedded Product Power-ups, and a forthcoming Digital Workforce of AI agents.

The new features, the company said in its announcement, “will give SMBs and mid-market companies a competitive advantage to scale and shift business dynamics without increasing resources, and enable enterprise and Fortune 500 companies to accelerate processes often slowed by scale.”

“Our ambition is to make monday.com an AI-first platform, where AI isn’t just an add-on but a core part of how businesses operate,” Or Fridman, AI product group lead, told Computerworld. “We believe AI has the potential to solve some of the most demanding business challenges, whether it’s making projects more predictable, improving decision-making, or automating complex workflows.”

AI Blocks in action

With AI Blocks, available in Monday Pro and Enterprise plans, users can set up their Monday board to automatically use AI on their behalf. Actions are customizable to specific projects and support both new and existing workflows without the need for technical expertise.

“It doesn’t require any prompting knowledge, and in less than 30 seconds, customers can set up their first AI Block,” said Fridman. “Ease of use is a major differentiating factor for us. Our platform is built on building blocks, helping people use technology easily.”

Each AI Block wraps an AI capability with context from a user’s work data. Customers choose the input (emails, documents, board column data), identify what they want the AI to do, and dictate where the output should be located (in a document or on a board column). The AI can be instructed, for instance, to prioritize tasks, assign workers to projects, extract information, assign labels, translate and improve text, and summarize updates.

“Monday.com’s AI Blocks are designed to integrate AI capabilities into existing processes without causing a lot of disruption,” said Melody Brue, VP and principal analyst at Moor Insights & Strategy. “The goal is to make AI more accessible to users who may not have specialized technical knowledge.”

Some use cases for AI Blocks, the company said, include lead categorization and file extraction (such as in HR). 

For example, a hiring manager could create a board to identify qualified candidates who should be moved on to the interview phase. She can add an AI-powered column with the task “assign labels,” which, based on her specific criteria, such as how many years of experience, skills and competencies a candidate must have, can automatically detect hireability. AI will then autofill columns with information from applicable candidate resumés, including email, phone number, and current employment, so that the recruiter can start making phone calls.

“AI Blocks provide the most help in aggregating information for reporting, reducing the time spent looking for relevant information, and then offering suggestions for the end user to communicate,” said Margo S. Visitacion, Forrester VP and principal analyst.

She emphasized, “users still need to verify the information is correct, but it can trim a lot of time from those tasks.”

Product Power-ups and Digital Workforce

Along with AI Blocks, Monday’s portfolio now has AI-generated Product Power-ups embedded throughout, which can automatically, for instance, rank best skill matches for tasks and reassign them accordingly, or identify project issues that require attention, such as resource scheduling or conflicts that might be causing delays.

“Our AI can proactively surface risks in large projects, making them more predictable and manageable,” said Fridman.

Finally, Monday’s upcoming “Digital Workforce” is a group of autonomously-operating AI agents that can perform tasks for users. The company will roll out its first agent, a “monday expert,” in March. This agent is designed to assist with onboarding new users and provide guidance around Monday features. Other planned agents include “deal facilitator” and “service analyzer.”

Fridman explained that customers can choose the relevant digital worker from the Monday marketplace and interact with it through chat and other in-product experiences. “It’s important to us to give value as quickly as possible without a long onboarding experience,” he said, pointing out that the AI workers will quickly learn, and adjust to, user preferences.

For instance, in customer support, an “AI-first digital service worker” could automatically resolve tickets so that human agents can move onto something more complex.

“If they work the way they are intended, they’ll make automating repeatable activities an integrated part of how an information worker executes their day job,” said Forrester’s Visitacion, “which can cut down on duplication of activities, task switching or potentially incorrect aggregation.”

Addressing concerns around predictability, reliability, safety

All told, the potential for AI agents is “enormous” when it comes to worker productivity, said Visitacion. “Companies can ask more complex questions and get meatier answers to support decision-making or automate the right workflows,” she said.

But she emphasized: “To get there, however, companies really must focus on structuring data and ensuring security and trust to ensure reliability.”

That’s a big, overarching concern: Addressing AI predictability, reliability and safety. Fridman said that Monday’s AI features follow the same data residency standards that exist across its portfolio, including multi-region support and encryption to ensure the privacy and security of customer data.

Further, “we fine-tune and optimize each AI engine and AI Block using proprietary techniques to ensure high quality, accuracy and built-in safeguards.”

Ultimately, agents will enable businesses to operate at a “previously impossible level, regardless of size,” Fridman said. These new capabilities are a “game-changer,” he said, particularly for SMBs that often lack the resources to scale like their larger enterprise counterparts. For instance, instead of hiring additional staff, SMBs can ‘hire’ digital workers to complete time-consuming, repetitive tasks such as managing operations, automating client communications, or streamlining order fulfillment.

“This allows them to scale faster, serve more customers, and focus on growth, all without the overhead of expanding their teams too quickly,” said Fridman. He added: “Our mission is to democratize AI, making it accessible and impactful for every business, not just the big players.”

Fridman emphasized that in 2025 and beyond, Monday is doubling down on AI-driven automation and intelligence.

“We want businesses to use AI not just for small tasks, but to redefine how work gets done,” he said. “That means evolving our digital workforce to handle increasingly sophisticated workflows. Our vision is a future where any company can build, automate, and optimize their operations using AI without needing data scientists or complex setups.”

In its announcement, Monday said, “To ensure AI remains accessible, monday.com offers a flexible and transparent pricing model for AI Blocks. Every plan includes 500 free AI Credits per month, providing teams with a simple way to explore the power of AI. For organizations with more significant needs, additional credits are available through buckets that scale with usage. Options range from a starter pack of 2,500 credits, geared towards lower usage, to enterprise buckets of 250,000 credits, providing flexibility for businesses of all sizes.”

AI ESM now GA

Monday also announced that its AI-first enterprise service management (ESM) platform is now generally available to all customers.

Just emerged from beta, the ESM offers AI-powered ticket resolution and automatic ticket classification and routing. It also features comprehensive service team dashboards intended to provide real-time insights into ticket trends, service performance and organization needs. Finally, a customizable customer portal allows users to access self-service options, submit tickets, track status, and communicate with the service team. 

Kategorie: Hacking & Security

Paris AI Action Summit: US and UK refuse to sign accord

11 Únor, 2025 - 21:17

The escalating electricity demands of artificial intelligence systems are raising concerns about the technology’s sustainability — but that’s apparently of little concern to the governments of the US and the UK.

They were among the invitees at the Paris AI Action Summit that refused to sign the “Statement on Inclusive and Sustainable Artificial Intelligence for People and the Planet,” the summit’s final declaration. The statement did win the approval of 58 countries, including China and India, and two supranational groups, the 27-member European Union (EU) and the 55-member African Union.

That’s more than signed the Bletchley Declaration by countries attending the AI Safety Summit organized by the UK in November 2023. The US and UK did sign that, as did the EU, China, and India, among others.

Signatories of the Paris summit statement agreed on six priorities:

  • Promoting AI accessibility to reduce digital divides
  • Ensuring AI is open, inclusive, transparent, ethical, safe, secure, and trustworthy, taking into account international frameworks for all
  • Making innovation in AI thrive by enabling conditions for its development and avoiding market concentration driving industrial recovery and development
  • Encouraging AI deployment that positively shapes the future of work and labor markets and delivers opportunity for sustainable growth
  • Making AI sustainable for people and the planet
  • Reinforcing international cooperation to promote coordination in international governance
Inclusion excluded

The US refusal to sign was likely triggered by the second priority of making AI inclusive: President Trump has ordered his administration to eliminate any reference to diversity, equity, and inclusion (DEI) from government websites.

But safety and sustainability are also not acceptable goals for the US, according to Vice President JD Vance, who addressed the summit on Tuesday morning.

“We stand now at the frontier of an AI industry that is hungry for reliable power and high-quality semiconductors,” Vance said. “If too many of our friends are deindustrializing on the one hand and chasing reliable power out of their nations and off their grids with the other, the AI future is not going to be won by handwringing about safety.”

Vance’s remarks about chasing out reliable power are likely a reference to moves in Europe to reduce reliance on electricity generated by burning oil and gas, European supplies of which have been disrupted by Russia’s invasion of Ukraine, in favor of renewable but weather-dependent sources such as solar- or wind-powered systems.

Coordination in AI governance is also going to be a point of contention. Even as the EU AI Act’s provisions begin to enter force, Vance warned summit attendees that “Excessive regulation in the AI sector could kill a transformative industry just as it’s taking off. The US, he said, “will make every effort to encourage pro-growth AI policies, and I’d like to see that deregulatory flavor making its way into a lot of the conversations at this conference.”

According to the BBC, the UK government also cited “global governance,” along with national security concerns, as reasons it refused to sign the Paris summit’s declaration.

America first

Vance was clear that his top priority is not accessibility or inclusion, but the US.

“This administration will ensure that American AI technology continues to be the gold standard worldwide, and that we are the partner of choice for others, foreign countries and certainly businesses as they expand their own use of AI,” he said.

But access to that technology will not be open to all.

“Some authoritarian regimes have stolen and used AI to strengthen their military, intelligence, and surveillance capabilities; capture foreign data; and create propaganda to undermine other nations’ national security,” Vance told summit attendees, adding, “This administration will block such efforts. We will safeguard American AI and chip technologies from theft and misuse, work with our allies and partners to strengthen and extend these protections, and close pathways to adversaries attaining AI capabilities that threaten all of our people.”

Billions in funding

Shortly after Trump’s inauguration, he announced that US AI companies would invest $500 billion in Project Stargate, designed to ramp up AI infrastructure in the US — although even with support from investors in Japan and the United Arab Emirates, barely a quarter of that sum is committed so far.

Vance predicted that investment would continue apace: “Of the $700 billion, give or take, that is estimated to be spent on AI in 2028, over half of it will likely be invested in the US,” he said.

But the US doesn’t have a monopoly on big projects. At the Paris summit, European Commission President Ursula Von der Leyen announced the EU’s intention to mobilize €200 billion ($207 billion) in investment in AI.

There’s some sleight of hand going on there too: While Von der Leyen talks of “mobilizing” €200 billion, only €20 billion of that is public money, and she’s expecting private enterprise to make up the rest.

Kategorie: Hacking & Security

An AI agent could help you buy your next car

11 Únor, 2025 - 19:59

Capital One has launched an AI agent designed to help customers with one of the more difficult and confusing purchase decisions: buying a car.

The new chatbot, called Chat Concierge, will help customers with everything from researching vehicles and scheduling test drives, to exploring financing options. The generative AI-powered assistant, one of many such projects at the financial institution, simplifies car buying by answering basic questions online with no dealership visit needed. It then directs them to existing online services.

Although Capital One’s auto loans are its smallest loan business, they still account for about 28% of its business, or $75 billion.

Chat Concierge is considered a customer service chatbot — a generative AI (genAI) automation tool that can handle simple user questions. The new service stands in contrast to Capital One’s own study last fall that found the in-person dealership experience remains vital for car buyers, even when they use digital tools to streamline early stages of the process. The report showed 88% of car buyers conduct at least half of the car buying process in person; 60% of buyers said sales reps contribute to trust.

“Car buyers’ trust in dealers is a key indicator of how transparent they perceive the car buying process — even with access to digital tools to complete key elements of their purchase,” the study concluded.

Even so, Sanjiv Yajnik, president of Financial Services at Capital One, said Chat Concierge will drive the future of car buying. “By leveraging our own internally developed AI tools to provide personalized, efficient, and transparent interactions, Capital One is reimagining car buying and setting a new standard for customer experience in the automotive industry,” Yajnik said in a statement.

Capital One’s AI assistant is part of a larger trend of companies deploying AI agents to tackle tasks often performed by entry-level employees, or to create efficiencies for high-level workers.

In the simplest sense, an AI agent is the combination of a large language model (LLM) and a traditional software application that can act independently to complete a task. The most basic AI agents include Chatbots such as OpenAI’s ChatGPT, Microsoft’s CoPilot, and Google Bard; they can answer user questions on a myriad of topics. AI agents can also act as spam filters, such as email spam detectors that use keyword matching and smart devices such as Thermostats that can follow set rules for raising or lowering temperature based on environmental conditions.

As AI-powered agents improve, they enable more personalized and effective customer service than early chatbots could deliver. Banks are using the genAI tools to resolve complex issues, setting new standards for efficiency. By leveraging customer data, AI assistants provide 24/7 support, handling thousands of inquiries at once, according to Arthur O’Connor, academic director of data science at the City University of New York (CUNY) School of Professional Studies.

“One of the most interesting developments is emotion recognition (ER), an emerging technology enabling chat bots to detect and respond to customer emotions, allowing for more empathetic and effective interactions, and thus engender customer satisfaction and loyalty,” O’Connor said.

Last month, Google DeepMind announced Project Astra, a research initiative aimed at developing a universal AI assistant that can process text, images, video, and audio inputs, enabling more natural and context-aware interactions. A key feature of Project Astra is its multimodal capabilities, allowing users to engage through various means such as speaking, showing images, or sharing videos. The assistant can remember details from past conversations and utilize tools such as Google Search, Maps, and Lens to provide informed responses.

The US Airforce recently announced it’s experimenting with a chatbot called NIPRGPT that will allow service members to engage in human-like conversations to complete various tasks, including drafting correspondence, preparing background papers, and assisting with coding.

Many AI agents will be integrated into existing software applications without users even knowing it. For example, Google Maps Navigation uses an AI model combined with traffic data and predicted conditions to provide the best route for drivers. Virtual Personal Assistants, such as Apple’s Siri, Amazon’s Alexa, or Google Assistant, use agents to predict user needs.

There are also learning AI agents whose algorithms are sophisticated enough to improve performance based on past experiences. Those systems include consumer recommendation services used on Netflix, Spotify, and YouTube, which all rely on AI to learn user preferences.

Agents that can become “smarter” include DeepMind’s AlphaGo, which learns and adapts to play the boardgame Go at a superhuman level.

Capital One’s Chat Concierge uses multiple AI agents that collaborate to mimic human reasoning. Instead of just providing information, the agents take action based on the user’s requests. They understand natural language, create action plans, validate them to avoid mistakes, and explain everything to the user, according to the bank.

For example, if a buyer asks for a list of trucks and then requests a test drive of the least expensive option, Chat Concierge can handle both tasks seamlessly. Concierge will also:

  • Simulate and validate plans to ensure they meet the car buyer’s needs and business policies.
  • Generate and deliver clear, natural language explanations of all the steps to the car buyer.
  • Let car buyers explore financing without leaving the dealer’s website.
  • Connect buyers directly to dealers through dealer websites, a navigator platform, and customer relationship management (CRM) apps, integrating customer info into the dealer’s CRM.
  • Work seamlessly with both Capital One and non-Capital One products.

“Capital One has a long history of using data, technology, and analytics to deliver superior financial services products and services for millions of customers,” said Prem Natarajan, chief scientist and head of enterprise AI at Capital One. “The launch of Chat Concierge is a key milestone in our customer-centered AI journey as we continue to focus on solving some of the most challenging problems in finance with technology.”

Kategorie: Hacking & Security

EU seeks to invest €200 billion in AI

11 Únor, 2025 - 19:01

The European Commission announced the mobilization of €200 billion (about $207 billion) for the InvestAI plan at the AI Action Summit in Paris on Tuesday, with the aim of enabling “open and collaborative development” of artificial intelligence in Europe. This was announced by Commission President Ursula Von der Leyen, who has also opened a new EU fund of €20 billion for AI gigafactories.

The strategy will thus finance four future AI gigafactories in the European Union (EU), which will specialize in training the largest and most complex models. These facilities will have around 100,000 state-of-the-art chips, approximately four times more than the centers currently under construction.

It is intended that companies of all sizes will have access to this computing power. These will have a focus on complex industrial and “mission critical” applications. Initial funding will come from different schemes, such as the Digital Europe Program and Horizon Europe and InvestEU.

The Commission already announced the first seven AI factories in December and will soon follow with the next five, which will represent the largest public investment in AI in the world and, it hopes, will unlock more than 10 times the amount in private investment.

A European AI Research Council will also be set up. Von der Leyen said, “We want AI to be a positive and growth force. We are doing this through our European approach, based on openness, cooperation and excellent talent. But we still need to leverage it. That’s why this unique public-private partnership, similar to a CERN for AI, will enable all our scientists and companies, not just the biggest ones, to develop the cutting-edge large-scale models needed to make Europe an AI continent.”

Kategorie: Hacking & Security

The Brave browser gets built-in functionality to run custom scripts

11 Únor, 2025 - 18:28

It’s been possible for a while now to modify web pages using popular extensions such as Tampermonkey and Greasemonkey, which can be useful for avoiding annoying ads or tracking attempts.

Now, starting with version 1.75 of the Brave browser, you don’t have to download this kind of add-on — because the feature is already built in. According to Bleeping Computer, the new feature can be used for everything from adding support for keyboard shortcuts to stopping the automatic playback of videos.

Information on how to write your own scripts is available on the Brave website.

Kategorie: Hacking & Security