Agregátor RSS

CEO of India's KiranaPro, which brings convenience stores online, vows to name the perp

The CEO of Indian grocery ordering app KiranaPro has claimed an attacker deleted its GitHub and AWS resources in a targeted and deliberate attack and vowed to name the perpetrator.…

It’s the groundbreaking therapy’s latest foray into battling solid tumors.

“Your cancer has spread” is terrifying news to hear. But it’s unfortunately common for people with colorectal cancer. The cancer is the third most common tumor globally and accounted for 930,000 deaths in 2020. One estimate suggests the disease could take up to 1.6 million lives in 2040.

Patients often die after colorectal cancer spreads to the liver, which makes removal extremely difficult. After the cancer metastasizes, treatment options are limited. Multiple rounds of full-body chemotherapy are the norm, but the therapy has severe side effects.

Patients have a survival rate of about 30 percent after five years, wrote Monica Casucci at IRCCS San Raffaele Scientific Institute and colleagues in a new paper.

The team’s research describes a more efficient, less toxic approach: CAR T therapy. Here, a patient’s own immune cells are extracted and fitted with proteins that enhance their ability to search and destroy cancers. The FDA first approved the revolutionary therapy in 2017 for people with a type of leukemia. Since then, six treatments have been approved for other blood cancers.

Casucci and her team have long sought to tackle metastasized colorectal cancer with CAR T. Compared to cancerous blood cells that naturally circulate in the bloodstream, solid tumors—for example, those in the colon, liver, or brain—are much harder to reach. The engineered cells must be directly infused into tumors with invasive surgery, an approach that’s even harder after a cancer has spread.

The new study aimed to treat solid tumors like blood cancer—with a single injection into a patient’s vein. The team engineered CAR T cells that could hunt down metastasized cancer cells. When infused into the veins of mice they found the engineered cells rapidly shrank tumors in the liver and large intestines without causing dangerous immune side effects.

The results “pave the way for a…clinical trial,” wrote the team.

Out-and-In

Our immune system already surveils cancer cells and sends T cells to destroy them before they expand. But cancers are tricky and rapidly mutate to evade the body’s immune defenses.

CAR T therapy uses genetic engineering to give natural T cells a boost.

Here’s how it usually works. Physicians first isolate T cells from a blood draw. They then insert genes encoding an extra “hook” protein that sits on the surface of the cells. This protein hook helps the cells locate and latch onto targeted cancer cells. Once infused back into the body, these superpowered cells are better at grabbing onto and destroying the cancer.

Success relies heavily on the hook’s design. These synthetic proteins—called CAR for chimeric antigen receptor—are designed to grab onto a specific cancer cell while ignoring healthy ones.

The first step is to find a protein target that’s unique to a type of cancer. Like all cells, the surface of cancer are dotted with proteins. These proteins form a sort of fingerprint.

Most blood cancers have similar fingerprints. But solid tumors are mashups of multiple cell types, each with its own signature, making it difficult to engineer targeted immune cells. These tumors can have attributes similar to healthy cells, wrote the team, meaning engineered T cells could inadvertently attack and cripple normal organs. Possible side effects also include cytokine release syndrome, where the immune system pumps out dangerously high levels of inflammatory molecules. In rare cases, the condition can be fatal.

Designer Missile

The team got to work with one goal in mind: Find a protein target that’s efficient and safe. After screening the genetic profiles of metastasized colorectal tumors from patients and scouring multiple protein databases, they landed on a protein called CDH17. Several gastrointestinal cancers—including colorectal cancers—express more of the protein than healthy surrounding tissues. Next, the team designed six versions of CAR T with protein hooks tailored to CDH17.

You can imagine these hooks as wobbly Lego structures with multiple sections. Some sections tunnel through the membrane of the immune cell. Others, on the outside of the cell, include the “hook” and a “hinge” that allows the protein to stretch, move, and flex so it can better detect and grab onto cancer cells. Yet another component anchors the protein to its host cell and, once a cancer cell has been found, sends signals inside that trigger its own cell to attack.

Two of their CAR T versions outperformed in tests. The team injected both either into the bloodstream or directly into the livers of mice with cancer that had metastasized.

The cells thrived and were roughly equally matched at fighting off cancer cells over a few weeks. Direct injection into the liver cleared out tumor cells faster, but the treatment was far more toxic compared to injection into the bloodstream. The mice experienced “irreversible weight loss,” wrote the authors, and all eventually died.

“Because rapid weight loss and fatal events have been recognized as clinical signs of CRS [cytokine release syndrome], we reasoned the toxicity observed might be” because of an overhyped immune response, wrote the team.

They were right. There was a spike in multiple inflammatory molecules when the CAR T cells were injected into the liver compared to a vein. The latter jab didn’t comprise the treatment’s efficacy and lowered the chances of a dangerous immune reaction.

Mice to Men

Both of the team’s leading CAR T therapies also worked in human tissues. In one test, the team transformed T cells from patients with advanced colorectal cancer that had spread to the liver into CAR T cells. They then made 3D mini-cancers, or cancer organoids, from the patients’ liver tumors. Both therapies grew in petri dishes and reduced the size of the organoids.

The CAR T cells ignored healthy intestinal cells, even when some also had a sprinkling of CDH17 on their surfaces. This is partly because the protein is nestled down into areas where healthy cells connect to each other, making it hard for CAR T to grab onto. In contrast, the protein is out in the open on the surface of colorectal cancer cells making them easier targets.

Although the study was only in mice and lasted a few weeks, it adds momentum to CAR T therapy for solid tumors. Another treatment for throat and stomach cancer is already in a phase 2 trial with promising initial results: The first phase increased survival rates compared to existing medications, although only for a few months. Many other similar trials are in the works.

The post CAR T Therapy Wipes Out Deadly Metastasized Cancer in Mice appeared first on SingularityHub.

Zuckercorp and Yandex used localhost loophole to tie browser data to app users, say boffins

Security researchers say Meta and Yandex used native Android apps to listen on localhost ports, allowing them to link web browsing data to user identities and bypass typical privacy protections.…

Microsoft, CrowdStrike, and pals promise clarity on cybercrew naming, deliver alias salad instead

Opinion  Microsoft and CrowdStrike made a lot of noise on Monday about teaming up with other threat-intel outfits to "bring clarity to threat-actor naming."…

The Google Chrome team issued an update to fix a high-severity vulnerability that is being actively exploited in the wild. The issue was also mitigated by a configuration change pushed out last Thursday to users of the stable Chrome version, which didn’t require a browser update.

Google Chrome exploits are highly valuable commodities on the black and gray markets with prices reaching hundreds of thousands of dollars. That’s because Chrome is one of the most hardened browsers and it uses process sandboxes to add additional hurdles to attackers. Bypassing all those defenses and achieving remote code execution on a system through Chrome usually requires chaining multiple vulnerabilities together.

[ See also: “Top 7 zero-day exploitation trends of 2024” ]

Third Chrome zero-day this year

That said, CVE-2025-5419, which was patched Monday in Chrome 137.0.7151.68/.69 for Windows and Mac and 137.0.7151.68 for Linux, is the third zero-day flaw fixed in Chrome this year. The other two, CVE-2025-2783 and CVE-2025-4664, were fixed in March and May, respectively. This highlights the elevated interest that hackers have in compromising Chrome users, despite the difficulty.

The new flaw was reported to the Chrome team by members of Google’s Threat Analysis Group, which is primarily responsible for defending Google infrastructure and users against government-backed attacks. This suggests the vulnerability was likely discovered in the wild, though details haven’t been released yet.

The vulnerability is rated as high severity, which indicates it can’t lead to remote code execution on the underlying OS on its own and likely must be combined with another flaw to do so. Otherwise, the flaw would have been rated critical.

Vulnerability in the JavaScript engine

The Chrome team described the vulnerability as an out of bounds memory read and write in V8, which is Chrome’s JavaScript and WebAssembly engine. The open-source V8 engine is used in other projects as well, including the Node.js runtime. Because the engine is designed to interpret and execute JavaScript and WebAssembly code, the vulnerability can likely be triggered remotely by users simply visiting web pages that load maliciously crafted code.

“Access to bug details and links may be kept restricted until a majority of users are updated with a fix,” Google said in its advisory. “We will also retain restrictions if the bug exists in a third-party library that other projects similarly depend on, but haven’t yet fixed.”

Aside from CVE-2025-5419, the new Chrome update also fixes a medium-severity use-after-free memory bug in Blink, the browser’s rendering engine. This vulnerability was privately reported by a researcher who received a $1,000 bounty for it.

The Chrome browser has an automatic update mechanism, but users who haven’t received it yet and want to prompt the update manually can access the Help > About Google Chrome menu to trigger an update check.

Partners banka od srpna chystá novinky. Vyčetli jsme je v nových obchodních podmínkách, které jsme prošli za vás.

Dnešní článek je spíše krátkou reportáží o nečekaném přírůstku do sbírky a jeho cestě ze stodoly až do vitríny. Jelikož se jedná o stejný hardware, který jsem zde již popisoval, budu se věnovat hlavně stavu nálezu.

Sonda do světa otevřeného softwaru. Dnes si představíme editor EPUB XHTML, vyzkoušíme grafické rozhraní pro nastavení touchpadů v Linuxu, řekneme si o emulátoru x86–64 na RISC-V a upravíme si pár PDF souborů.

Předčasně zveřejněná recenze Radeonu RX 9060 XT poskytuje první pohled na celkový výkon nového mainstreamu od AMD. V podstatě nepřekvapuje a naplňuje očekávání…

Muzeum je to pěkné, příjemné a ilustrativní, rozhodně stojí za prohlídku. Přivítá nás strašlivá Gorgó (Gorgóna). K tomu pár poznámek o kykladském muzejnictví.

Astrobiologové říkají, že Saturnův Enceladus nejspíš alespoň teoreticky splňuje podmínky, které by umožnily výskyt organismů pozemského typu. Máme v podstatě dvě možnosti. Můžeme tam pátrat po původních organismech, v podstatě navěky, protože stále bude zbývat nějaká lokalita nebo metoda, co jsme ještě nezkusili. Anebo tam pošleme pozemské organismy a budeme sledovat, co se stane.

…aneb Třicáté výročí popisu obřího teropoda z Argentiny

Startup Hermeus se dere vpřed v legendárním závodě dneška o hypersonické technologie. Vsadili na specializované prototypy menších velikostí a jak se zdá, svědčí jim to. Jejich Quaterhorse Mark 1 se vznesl za necelý rok od návrhu designu. Už příští rok chtějí dosáhnout s prototypem Mark 3 rychlosti Mach 5.

OpenAI's next big foundational model is GPT-5, and the AI startup is hoping that the model will compete a little more with rivals. [...]

ChatGPT's memory feature is now better and capable of referencing past conversations for free accounts. [...]

Hewlett Packard Enterprise (HPE) has issued a security bulletin to warn about eight vulnerabilities impacting StoreOnce, its disk-based backup and deduplication solution. [...]

Generative AI (genAI) is enhancing worker value and productivity, not replacing people — and that’s true even for roles that are vulnerable to automation, according to new new research by PricewaterhouseCoopers (PwC).

Analysis by PwC of nearly 1 billion job ads throughout the world showed that genAI-exposed industries have tripled revenue per worker since 2022, proving genAI investments are paying off. Overall, the report showed that AI is transforming jobs, boosting productivity, wages, and skill demands, rather than causing widespread job losses.

The report flies in the face of comments by Anthropic CEO Dario Amodei, who told Axios AI could wipe out half of all entry-level white-collar jobs — and spike unemployment to between 10% and 20% in the next one to five years. Anthropic also just made generally available a version of its Claude AI assistant (Claude Code) that can write, edit and debug code, making it nearly as good as a human developer. Known as “vibe coding,” the use of natural language to develop software is expected to boom over the next few years.

Last month, Microsoft CEO Satya Nadella said AI now writes up to 30% of the company’s code, and that’s expected to quickly increase.

A report from MIT Technology Review Insights found that 94% of business leaders now use genAI in software development, with 82% applying it in multiple stages — and 26% saying they used it in four or more.

“Software developers are evolving into strategic technology orchestrators who harness AI to drive unprecedented business value,” said Kye Mitchell, head of tech staffing firm Experis North America.

The impact of genAI on hiring has been stark, as companies grapple with cleaning, organizing, and sharing data stores for potential use by the technology. Demand for database architects skyrocketed, leaping 2312% in the past year, Mitchell said. Jobs for statisticians also rose sharply, up 382% in the same time frame.

“AI isn’t replacing jobs — it’s fundamentally redefining how work gets done,” she said. “The break point where technology truly displaces a position is when roughly 80% of tasks can be fully automated. We’re nowhere near that threshold for most roles. Instead, we’re seeing AI augment skillsets and make professionals more capable, faster, and able to focus on higher-value work.”

Not surprisingly, PwC also found AI use increasing across all industries, including traditionally low-tech ones like mining and agriculture. The firm also found that wages in AI-exposed sectors are rising twice as fast as in less-exposed sectors. Workers with AI skills earn a 56% wage premium, up from 25% last year.

And AI-driven changes to worker skills are accelerating, with a 66% faster shift in skill requirements in AI-exposed roles.

AI job postings continue to rise, despite a softer job market, showing persistent demand for talent. The information and communication sector leads AI skill demand, while construction and healthcare lag behind, PwC reported.

In the US, jobs with high AI exposure have seen a slowdown in job postings between 2019m and 2024, but greater skill evolution, highlighting how “AI reshapes roles more than it eliminates them,” PwC said.

Employers have continued to pursue skills-based hiring strategies over the past three or so years. About one-half of all April tech job postings did not specify a need for a four-year academic degree, according to CompTIA, a nonprofit trade association that issues professional IT certifications.

Jobs with high gen AI exposure in the US have seen a decrease in degree requirements, falling from 63% in 2019 to 53% in 2024. Jobs exposed to automation now require degrees less often today (41%) than they did in 2019 (56%).

While degree requirements in white collar job listings have markedly decreased over the last several years, the shift toward more AI-based job roles has affected employment – especially in IT-related positions. In April, the tech industry lost 214,000 positions as companies shifted toward AI roles and skills-based hiring amid economic uncertainty, according to an evaluation of the US Bureau of Labor Statistic’s latest jobs report.

TAG team spotted the V8 bug first, so you can bet nation-states weren’t far behind

Google revealed Monday that it had quietly deployed a configuration change last week to block active exploitation of a Chrome zero-day.…

It has taken just three years for the GenAI generation of AI to reach the level of use the Internet itself took 23 years to achieve, says legendary US investor Mary Meeker in her latest Trends report. 

That’s why, unless Apple has viable plans we don’t yet know about, it needs make an AI-related acquisition soon. It needs to do so because the new generation of AI is already achieving a global resonance we’ve never seen before. 

With the impact of generative AI (genAI) now spreading across tech, finance, social, politics, and employment, Apple needs to be part of the convergence to maintain relevance.

Where the puck is going

Meeker’s report gives you a solid sense of this, and in doing so shows the extent to which genAI is being deployed across developing economies in Africa, Asia, Latin America, and the Middle East.  That matters more because many of these areas have not enjoyed ready access to the internet before, which means they aren’t starting with Usenet and scaling to FaceTime – they are beginning their internet adventure with AI. These first-to-AI cohorts will soon become the first “AI-native” populations, driving economic growth in those geographies.

Speed and execution

This is a fast game – more Blink than Bridge. Meeker’s report points at the extent of this disruption. “Seem like change happening faster than ever?” it asks. “Yes, it is,” the report responds, providing a range of metrics to show it — not least the swiftness with which genAI has achieved 800 million weekly active users since October 2022. 

“Rapid advances in artificial intelligence, compute infrastructure, and global connectivity are fundamentally reshaping how work gets done, how capital is deployed, and how leadership is defined – across both companies and countries,” the report says.

Smarter than nothing

Apple, stung by slow development of Apple Intelligence, needs to maintain a place in the race — but the speed of this race underlines the huge risk the company has been forced to take as a result of its well-publicized AI failures.  Apple can’t keep making these errors. It should, perhaps, have been faster to embrace OpenAI when it emerged, rather than permitting Microsoft to get there first.

That error gave Microsoft Copilot wings Siri still can’t match.

Apple may be on the cusp of repeating that mistake with Samsung, which is allegedly looking to take a position with Perplexity. Apple is already working with Perplexity, but recent reports claim Samsung is preparing a wide-ranging deal to use Perplexity AI to provide search on Samsung smartphones. Some wire reports this morning suggest Apple is also interested in Perplexity, citing an older statement Eddy Cue last month made during his testimony at the Google Search trial: “We’ve been pretty impressed with what Perplexity has done, so we’ve started some discussions with them about what they’re doing.”

Grab your partners

The risk is that Perplexity goes with Samsung, leaving Apple in need of a strong AI partner. Apple’s approach might be to become polyamorous, with partnerships with OpenAI, Google Gemini, Perplexity, and others providing some of what its devices need to be part of the AI deployment party. That may even be enough, for a while.

But as competitors begin to chip away at the Android/Apple duopoly with their own alternative hardware capable of running AI (i.e., precisely the kind of hardware former Apple designer, Jony Ive is working on with OpenAI), Apple has too much to lose — far too much to lose.

Existential crisis

That’s why one recent leak claiming Apple’s management has adopted a “by any means necessary” approach to bringing its platforms up to speed for AI is reassuring. After all, it’s not such a huge step, once you accept the need for partnerships with AI service providers, to figure out that perhaps there’s a good reason to acquire one of those providers. 

Not only does Apple have the cash to do it, but just as its huge investment in processor maker PA Semi eventually drove decades of hardware design, so too will AI drive the coming decades in computing. It’s an existential necessity. 

But does Apple need to acquire one of the larger household names in AI?  Probably not. 

Raise them up

There are other firms, some small, some large, that may already have some of the tech that Apple needs. Many of these may lack the infrastructure to deliver their services at a big enough scale to meet the needs of Apple’s billion-plus users. 

Apple might be able to help with that. It has, after all, been making significant investment in Private Cloud Compute — to the extent we’ve even heard it has production lines churning out servers to support that service.

Why make so many servers? With 1 billion users, it might just be to support Apple Intelligence. It could also perhaps enable Apple to offer developers an AWS-style B2B service for secure and private AI. But it could also become an infrastructure on which to host any AI solution Apple might eventually acquire, enabling promising tech to swiftly reach an audience of millions at a time when AI adoption is absolutely spiking.

Will this happen? Even Bloomburg’s Mark Gurman doesn’t seem to know just yet.

Should it? Probably. 

You can follow me on social media! Join me on BlueSky,  LinkedIn, and Mastodon.