wiredmikey shares a report from SecurityWeek: Meta is the latest major AI developer to admit that its models broke loose during cybersecurity testing and hacked external systems. The tech giant said in a statement to the media on Wednesday that the incident occurred during independent evaluations conducted by Israeli AI security startup Irregular. The tested AI models were inadvertently allowed to access the internet due to a misconfiguration, which led them to exploit a vulnerability in an unnamed third-party service. It's unclear if it was a known flaw or a zero-day.

The Information [gated] learned that the Meta AI attacks involved the company's advanced Muse Spark 1.1 model, which breached an unnamed organization's systems and made unauthorized changes to its internal environment. Meta said it learned of the AI models going rogue after being notified by Irregular. The company is conducting an investigation and it has promised to issue a "full retrospective" once it has all the facts. A spokesperson for Irregular said the incident was the "exact same evaluation-environment issue that was already disclosed by Anthropic last week" and that it did "not involve a "sandbox escape or a sophisticated cyber action."

It contrasts with OpenAI, whose AI agent independently exploited a novel vulnerability to reach the internet during cyber testing. Not only did it breach Hugging Face but it also hacked multiple third-party accounts and services as part of the attack.

An anonymous reader quotes a report from Ars Technica: Routine cybersecurity testing of frontier AI models sparked a series of unexpected security incidents -- the most serious case arising when Anthropic's Mythos 5 model attempted to insert malicious code into an open source software application and created fake identities to deceive the human developers maintaining the project. The security incidents occurred during a cyber evaluation of seven leading AI models' capabilities by the AI Security Institute (AISI), a research organization within the UK government, in late July. The researchers discovered (PDF) 19 instances in which "AI agents took unsanctioned action on the live Internet, including cases that targeted real people and organizations," according to an AISI blog post published on August 4.

Almost all the "autonomous, unsanctioned" actions came from Anthropic's Mythos 5 model, with two such actions coming from OpenAI's GPT-5.6 Sol. [...] The most serious case involved Mythos making multiple attempts to execute a supply chain attack on the open source project repository hosted on the developer platform GitHub, including using social engineering techniques to try to convince the repository's human maintainers to merge malicious code into the repository.

After first opening a pull request to merge the malicious code into the repository, Mythos created fake online "sock puppet" personas that claimed to have independently reviewed and verified the code as not containing malware. The AI agent also sent five emails to two human maintainers of the repository, including some emails containing malware and others attempting to persuade a maintainer to accept the pull request. Mythos even opened a GitHub Issue on a second repository -- also owned by a maintainer of the first repository -- that contained a prompt injection with malicious instructions targeting "issue-triage AI coding agents." This line of attack came from Mythos reasoning that the repository maintainer could be an AI coding agent such as Claude Code.

Security researchers found that Apple's iCloud Private Relay can expose users' real IP addresses because some passkey-related requests bypass Safari and its proxy protections at the operating-system level. "In short: any website that supports, or pretends to support, passkeys can see the user's real IP address despite having iCloud Private Relay on," security researcher Tommy Mysk, who discovered the issue along with Talal Haj Bakry, told 404 Media. The flaws also affect OnionBrowser, an iOS app for browsing the web through the Tor anonymity network. It does not, however, impact the official Tor Browser itself. From the report: The researchers developed a site that lets Private Relay users check if the issues impact them. In 404 Media's tests, the site did return the real IP address of a user that was supposed to be protected by Private Relay.

[...] In a quirk of how passkeys work -- a broadly secure alternative to usernames and passwords which use the WebAuthn standard -- a user's device makes a web request outside of the browser itself. Meaning, that request essentially bypasses Private Relay and exposes a user's real IP address, even though to them it may look like they are simply interacting with a website as normal.

"Because the fetch is issued by the operating system's credential service rather than by Safari, it never enters Private Relay's proxied path. The destination server sees the device's real IP address either way," the researchers write in their research. [...] "We have already informed them. They said the issue was âdire,' but they let us disclose the issue. They didn't provide any time when they will address this," Mysk said.

Cloudflare has open-sourced Cloudflare OS, an Apache 2.0-licensed platform that lets organizations build AI agents, apps, and workflows using curated company data and tools within isolated, governed environments. Despite the name, it is not a traditional operating system but a framework for securely managing organizational AI workloads. Phoronix reports: Cloudflare OS is already used internally at Cloudflare and is described in today's announcement as: "Cloudflare OS starts with a conversation in your browser, like many other AI tools. What makes it different is that each conversation is grounded in the context and skills your organization has curated. Give your workspace a goal, and it can draw on that knowledge and work with the tools and data your organization already uses to achieve it.

Cloudflare OS combines three parts:
- An agent workspace grounded in context and skills your company curates, with an isolated runtime where agents can write and run code.
- A new security and governance framework for safe access to internal data and services.
- A platform for personal, modifiable apps that people can build, share, and continue changing.

What begins as a conversation can become a doc, an app, or a workflow that continues doing the work." You can learn more at os.cloudflare.app.

Apple has capped the number of open bug-bounty reports researchers can submit after being flooded with low-quality and sometimes entirely fabricated vulnerabilities generated by AI. MacRumors reports: The Financial Times learned of the limit after cybersecurity startup Bynario used ChatGPT to locate more than 50 macOS bugs in three weeks. Bynario found a privilege escalation exploit that could let an attacker get unrestricted access to a Mac, but was unable to report it because Apple limited the number of bug reports Bynario could submit. Bynario sent eight reports to Apple in 2025, and another five in 2026 before hitting a restriction.

Bynario's founder said it is a "very difficult time in the industry" because companies are being "flooded by the sheer amount of bugs." Apple has since been in contact with Bynario and is reviewing the company's submissions. While Apple now has a cap on the number of open submissions a researcher can have, researchers can request an increase to make sure Apple's security team doesn't miss a critical vulnerability.

An anonymous reader quotes a report from The Guardian: Apple has launched a new legal challenge against a UK government demand to access its customers' highly encrypted data, a year after the Home Office agreed to abandon its previous request. The US tech company launched the legal complaint last month at the Investigatory Powers Tribunal (IPT), an independent court that has the power to investigate claims that the UK intelligence services have acted unlawfully. The UK government had made a second request to Apple to grant it a "back door" to encrypted iCloud data belonging to British users, according to an order issued by the court.

Britain backed down on its original demand for access to data from UK and US customers last year, after a heated transatlantic tussle over encryption between London and Washington. UK authorities subsequently issued a new "technical capability notice" (TCN) to Apple that did not apply to American users. Apple is seeking to challenge the British government's powers to issue TCNs under the UK Investigatory Powers Act, according to the details of the new legal case first reported by the Financial Times. [...] The original TCN issued last year asked Apple for the right to see users' encrypted data protected by its advanced data protection (ADP) program in the event of a national security risk.

Apple said the removal of the tool -- which not even it can access -- would make users more vulnerable to data breaches from bad actors and other threats to customer privacy. Creating a "back door" would also mean all data was accessible by Apple, which it could be forced to share with law enforcement possessing a warrant. As a result, Apple withdrew UK customers' access to its ADP program in January 2025. The Home Office has maintained that the Investigatory Powers Act, under which such orders are issued, contains robust safeguards and is used only when absolutely necessary.

A Russia-linked group tracked as Midnight Blizzard has compromised hotel and conference Wi-Fi portals worldwide, redirecting guests to phishing pages and fake software updates that steal credentials, session tokens, and other sensitive data. Microsoft says the campaign, dubbed CaptiveCrunch, "targets traveling employees generally rather than a particular sector," reports iTNews. From the report: Midnight Blizzard, tracked internally by Microsoft under its earlier codename NOBELIUM, is attributed by the US and UK governments to Russia's SVR (Sluzhba Vneshney Razvedki) foreign intelligence service. Microsoft's technical analysis said compromises occurred in "several countries" without naming them, and it did not give a total number of affected venues, organisations or individuals.

A related investigation published earlier in July by security firm ReliaQuest, and which Microsoft cited in its report, found compromised captive portal gateways across multiple United States cities as well as in India and Saudi Arabia, mostly at hotels. ReliaQuest said the traffic it observed came from organizations across financial services, professional services, legal, health care, energy and retail, suggesting the campaign targets traveling employees generally rather than a particular sector.

[...] Where attackers gained a foothold, Microsoft said they deployed two main tools: CornFlake, a Windows remote access trojan (RAT) written in Go capable of keylogging, screenshot and webcam capture, audio surveillance and credential and session token theft. They would also drop ChocoShell, an in-memory PowerShell infostealer targeting browser cookies, saved passwords, Microsoft 365 single sign-on (SSO) tokens and wi-fi credentials. Microsoft also said it has seen indications the attackers might be targeting Android devices with similar prompts urging victims to download and install an APK file.

theodp writes: During Microsoft's 2026Q4 earnings call, CEO Microsoft Satya Nadella took time to tout a dashboard he personally created using AI from a Morgan Stanley analyst's PDF research report, which suggested a rosy payback for the so-called MAG7's ('Magnificent 7' companies) massive capital expenditures on AI (to which Nadella later added a "not financial advice" disclaimer). "It would be fun for you, Adam. I think one of your colleagues put out an ROIC [Return on Invested Capital] document. I took that document to Copilot, which is a PDF, and I said, 'Build me a new Power BI dashboard, essentially.' But here is the thing. It built a rich semantic model that went into my Fabric with OneLake that brought all the data in from the external sources. In fact, it was current with all the SEC filings of all the MAG7. And then on top of that, the repo itself is in GitHub, but the artifact is sitting in my Copilot as a site. That, to me, is a classic example of an enterprise-wide workflow. I, as a knowledge worker, could go create a dashboard. The data engineer can go to Fabric and find the artifact. The professional developer can go to the repo and find it in GitHub. And by the way, it's all registered with Agent 365. That's a little bit of what Amy is describing as the coming together of a new way to work, even while at the same time, bringing IT, security and manageability of it."

After Nadella's show-and-tell drew an underwhelming response during the call ("That's very helpful. Thank you." said the Morgan Stanley analyst whose team's work Nadella scraped with AI), Nadella turned to social media with posts on LinkedIn (12M followers) and Twitter/X (8M followers) to make the case for why his DIY project was such a brilliant demonstration of how AI enables governance, controls, security, development, testing, deployment, maintenance, data analysis/modeling, visualization, usability, and value. "Some more detail on the ROIC Intelligence App I built yesterday and mentioned on today's earnings call," Nadella wrote on LinkedIn. "I took the PDF that Brian Nowak at Morgan Stanley put together for Hyperscale ROIC this week and used Copilot code (coming in our new superapp) with a single prompt + skill (/drill-me) to create the plan, then used autopilot in auto to create the full app (with history, lookups, scenarios, what-ifs, etc). And /rubber-duck to test. And the best part is that all the artifacts are in my enterprise environment. My app is in Copilot, my code is in GitHub Enterprise; all my data pipelines/lake/semantic models are in Fabric. And everything is under Agent 365 IT/Sec/FinOps control! So this is not about Tokenmaxxing or vibe coding. Every step of the way the rails are engineered to create value, making everything a long-term reusable asset, with governance/security, and cost controls. This is the full system to drive business value. Disclosures: This is all pulled from public sources, and for illustrative purposes only...not financial advice! :) Here is the app and architecture..."

Not unexpectedly, the accompanying screenshot of a splash page for the BI app and a buzzword-laden complicated architecture diagram drew universal praise from LinkedIn fans, but also a few barbs from less-than-impressed commenters on Nadella's Twitter/X post, some of whom suggested Nadella's project might even represent a jump-the-shark moment for AI mania. "That he doesn't see whats wrong with saying 'My app is in Copilot, my code is in GitHub Enterprise; all my data pipelines/lake/semantic models are in Fabric' is exactly why MS is failing at AI,'" replied @PassingPixels on X/Twitter. "Dude is having to run 5 different systems to emulate babies first vibe code." @zigmund_ignatov added, "Why do we call glorified slide show an app?" @Mathupiriyan quipped, "Looks like Copilot just turned a PDF into a profit crystal ball." Unimpressed, @Markusndnb remarked, "So you created a web page using tons of proprietary MS tools." And @FishyAccounting called on Nadella to show-his-work, saying "Post the prompt or it didn't happen." (btw, Microsoft President Brad Smith similarly declined to provide the prompt for his own self-described amazing AI DIY reporting project that he touted at Microsoft's Shareholder Meeting last December).

So, does Nadella's self-promoted AI reworking of someone else's PDF research report strike you as an amazing example of everything that's good about AI, or does it conjure up memories of The Emperor's New Clothes?

An anonymous reader quotes a report from TechCrunch: Several popular Samsung smart TV apps contain code that share the owner's internet connection with strangers, potentially putting millions of Samsung smart TVs at risk of hijacking, according to new security research published on Monday. Some of these apps claim to have been installed on hundreds of millions of smart TVs in people's homes, per the app developers. At least one of the smart TV apps was a simple Pac-Man game that Samsung had endorsed and prominently featured in its "Editor's Choice" section on customers' TV screens. These apps contain software that funnels outsiders' web traffic through ordinary home and office internet connections, known as residential proxy networks (or "resproxies"), which are increasingly being linked to cybercrime. When opened, apps with resproxy code can turn the smart TV into an always-on tunnel for outsiders to funnel their web traffic through, known as an exit node -- even when the app is no longer open.

The security research by Norwegian cybersecurity company Mnemonic describes a perfect storm of problems that allows low-quality apps to proliferate across Samsung's app store, containing code that puts users at risk of having their internet connections tapped by a rogue app. Many of these apps are bare-bone shells, made from only a few lines of code, and are designed solely to load content from another website, such as a game. While such smart TV apps load content from another server, any review of these apps sees only the few lines of code within, and not necessarily the content itself. "What was reviewed is not necessarily what is running," wrote Harrison Sand, an offensive security consultant at Mnemonic.

After TechCrunch contacted Samsung with a request for comment about the research, the electronics giant said in an emailed statement that it was banning apps that share their users' internet connections, and will remove apps that contain the functionality. "We have already restricted new app registrations that incorporate such proxy functionalities on our Smart TV platform," said a Samsung spokesperson. "We are currently implementing strict platform-wide developer policies explicitly banning residential proxy SDKs, and we are working to identify and remove all apps currently available in our store that contain these components." LG also recently announced plans to suspend apps containing ResProxy software after a security firm found that roughly 42% of apps in its TV app store allowed unknown third parties to route internet traffic through users' televisions without their knowledge.

Slashdot reader prisoninmate shares this report from 9to5Linux: Coming ten days after the previous Linux kernel security update, which only fixed 12 vulnerabilities that may lead to a privilege escalation, denial of service, or information leaks, the new Debian 13 Linux kernel security update is a massive one, and it patches no less than 68 security vulnerabilities in the Linux 6.12 LTS kernel.

Debian 13 "Trixie" kernel security update are CVE-2026-64530, a use-after-free in the traffic-control subsystem leading to remote denial-of-service with potential for remote code execution, and CVE-2026-64531 (a.k.a. OVSwrap), a local-root vulnerability in the Open vSwitch datapath leading to local privilege escalation to root...

All Debian 13 "Trixie" users are urged to update their installations to Linux kernel 6.12.100-1 as soon as possible.

Even as Hollywood performers protest and Hollywood studios sue "in their war on AI," reports the Los Angeles Times, "the entertainment industry is deepening its dependence on it." Among hundreds of job postings in late June, more than one in 10 was likely connected to AI. The top studios' public postings suggest they have been recruiting people to build AI tools. They are also recruiting teams to defend their intellectual property against unauthorized AI use. "There are plenty of studios that are hiring [for AI] but never talk about it in public," said Yoland Yan, a co-founder of ComfyUI, a company that helps studios juggle different AI tools. Companies have been hesitant to detail how they use generative AI in film production — partly because they are concerned about consumer and union backlash. Some in Hollywood described AI use as the new cosmetic surgery, where everyone knows it is happening, but few will admit to it...

Although some companies may be shy about sharing their AI plans, big stars who don't have to answer to others have been more open about their embrace of the new technology for storytelling. Rejecting AI is like picking a horse and buggy over a car, said "Star Wars" creator George Lucas. "Artificial intelligence means it's much easier for us to make movies," he told a trade magazine earlier this year. "There's nothing you can do about it. That's progress. It's the future." Some in Hollywood have a softer stance on artificial intelligence, with studios cutting deals with AI companies, and filmmakers like Martin Scorsese backing AI companies. Ben Affleck launched an AI film tech company then sold it to Netflix for half a billion dollars. When launching InterPositive, Affleck said he wanted to keep "storytelling human" by building AI tools that could fix lighting, generate missing shots and other things while "keeping creative decisions in the hands of artists...."

Disney, Netflix and Amazon had job postings that were about using AI on the creative side of the business. Universal, Paramount, Warner Bros. and Sony had job ads suggesting they were also using AI but for marketing, distribution and audience analytics. The postings suggest the Disney, Netflix and Amazon studios are building repeatable AI workflows for visual effects, animation, sound and dubbing. The companies also seem to be building in-house teams to develop custom generative-AI models, while also using third-party software.

None of the jobs advertised were to create AI that wrote scripts or created AI actors.
Ironically, the Times used Claude Code to build a scraper to identify the job postings, their article acknowledges.

• Three Disney jobs were for "content security," assessing AI tools and guarding against piracy, watermarking and rights-protection work. But Disney is also hiring PhD-level talent "to study 'computer graphics and AI' for Pixar and Disney films," according to the article, and "people to 'bridge the gap between research and practical studio application.'"

• Disney-owned visual effects shops Industrial Light & Magic "was searching for supervisors to 'explore emerging technologies (including AI/Machine Learning)' to develop new production workflows."

• Audio post-production unit Skywalker Sound "seemed to be recruiting to build proprietary AI models for soundtracks, voice separation, and voice transfer, the process of taking a speaker's tone and pitch, and applying it to new content."

• Amazon "was hiring a principal AI executive to drive AI-tool adoption across production, plus roles in operations automation and LLM content classification."

Researchers found enterprises are spending time troubleshooting cloud connectivity due to increased AI workloads, reports Computer Weekly. More than 400 IT and infrastructure decision-makers (US and UK) were surveyed for internet/cloud/AI exchange operator DE-CIX by market researchers Censuswide. But despite 96% of respondents claiming their enterprise networks are ready for cloud/AI loads, the average IT team still spends more than 11 hours each week resolving cloud connectivity problems: Other leading concerns included downtime or reliability issues (26%), latency or slow performance (28%), and security vulnerabilities/DDoS attacks (27%). Cost of connectivity, staff expertise and lack of visibility/control over data flows were also cited as major challenges... As a result, as indicated in the study, many businesses are now turning to private interconnection, which enables enterprises to connect directly to cloud providers over dedicated infrastructure rather than routing traffic across the public Internet. Designed to deliver lower latency, greater resilience, enhanced security and more predictable performance, private interconnection has become an increasingly important way of supporting modern cloud and AI workloads. Specifically, the data showed that 61% of companies are already using private connectivity to clouds, while another 31% are actively considering it... [And 71% of enterprises with 1000 or more employees]

Only 8.62% of the smaller companies were spending 21 to 40 hours per week dealing with connectivity issues, while just 2.53% of the largest companies in the sample do. Summing up these findings, DE-CIX said that together they suggest direct interconnection is rapidly becoming a core component of enterprise cloud and AI infrastructure and a competitive advantage for companies, though optimising interconnection strategies clearly remains a pressing challenge for small and medium-sized enterprises... "Every AI application depends on data moving quickly, securely and predictably between users, clouds and AI infrastructure. Our research suggests that far too many enterprises are still spending valuable time trying to maintain that kind of connectivity, with more than a third spending between 11 and 20 hours per week, and just under one in 10 spending between 21 to 40 hours per week. This confirms what we already knew — that that network architecture can make or break AI adoption."
Elsewhere The Register reports that cloud infrastructure services "grew at their fastest for eight years during the second quarter of 2026, thanks to the AI craze and continued demand for flexible and scalable IT infrastructure." According to the latest figures from Synergy Research, enterprise spending on cloud infrastructure passed $143 billion in Q2, a year-on-year growth rate of 43 percent. This followed 11 successive quarters of increasing growth rates, during which the market has now doubled in size... "AI has, of course, driven most of that incremental growth, and we now see year-on-year growth rates of 165 percent for AI-specific cloud services...." And the top three global players continue to dominate the market, with Amazon Web Services (AWS), Microsoft Azure and Google Cloud together accounting for 67 percent of all the cloud revenue during the quarter. That percentage has increased since the third quarter of last year, when the triumvirate made up 63 percent of enterprise cloud infra spending.

New York has finalized new rules that will govern social media apps in the state starting on January 25, 2027. The law prohibits social media platforms from sending notifications to minors between midnight and 6 a.m. without parental consent. And minors "will only be shown content from other accounts they follow or otherwise select in a set sequence, such as chronological order," rather than "the default algorithmically personalized feeds... unless they get parental consent for an addictive feed." (Social media companies "must offer at least one alternative method for age assurance besides providing a government-issued ID," the announcements points out, and any information used to determine age "must not be used for any other purpose and must be deleted or de-identified immediately after its intended use.")

But meanwhile, the EFF writes that a committee in the U.S. Senate is considering the SCREEN ACT, "a sweeping age-verification bill that would require online services to verify users' ages before they can access any sexually explicit content. If this bill passes, it will force millions of adult internet users to give up their anonymity, privacy, and security before they access lawful speech." Unlike many state-age verification laws — which have been harmful in their own right — the SCREEN Act has no requirement that a significant portion of the website consist of sexually explicit content that is harmful to minors. The bill requires nearly any service hosting even a single piece of sexually explicit content to verify the ages of its users. The result is that the bill would apply not only to adult content sites like PornHub or OnlyFans, but also streaming services like Netflix, and social media platforms like Reddit, Discord, or Bluesky, if they host any adult content...

Under the SCREEN Act, the "bouncer" will be a digital age-verification service that captures your personal information and saves it to a database for an unspecified amount of time. The consequences of the bill won't be limited to minors. If websites and apps are expected to reliably identify teenagers, adults will be asked to prove they are adults. Even worse, the SCREEN Act is a privacy and data security nightmare. One provision of the bill requires services to take reasonable steps to protect the data collected and to not maintain for longer than is necessary. But these are terribly weak protections that impose no meaningful collection, use, or retention limits on services collecting people's private information...

The SCREEN Act also targets virtual private network (VPN) users and providers. The bill requires covered websites to verify users' ages based on their IP addresses unless the service can determine that the user is outside the United States, and specifically requires age verification on traffic coming from known VPN addresses. In practice, this discourages the use of VPNs and proxy servers, which millions of people rely on for legitimate purposes such as protecting personal privacy, securing public Wi-Fi connections, safeguarding journalists and activists, and preventing data tracking...

The SCREEN Act creates onerous age-verification rules that will block adults from accessing lawful speech, curtail their ability to be anonymous, and jeopardize the data security and privacy of all internet users.

"GitHub and the Python Package Index (PyPI) have introduced new policies meant to boost supply chain security," reports SecurityWeek, "by preventing the fast propagation of poisoned package versions and the poisoning of old and long-stable releases." To prevent the fast delivery of malicious code through the immediate fetching of brand-new releases, GitHub has introduced a Dependabot cooldown, where the automation tool waits for at least three days after a release has been published before opening a pull request. "Waiting a few days before adopting a new release gives maintainers, security researchers, and automated scanners time to spot a malicious version and get it pulled before it ever reaches your pull requests," GitHub explains.

The three-day cooldown only applies to non-security version bumps, and the behavior can be modified through the configuration option in the dependabot.yml. "Three days as the default balances two goals: it pushes you past the window where most of these attacks live, and it doesn't hold your dependencies back longer than necessary," GitHub notes.
And the Python Package Index (PyPI) "now rejects new files being uploaded to releases that are older than 14 days," according to a recernt blog post from the Python Software Foundation's security developer-in-residence Seth Larson: This restriction was put in place to prevent old and long-stable releases from being poisoned in case publishing tokens or workflows of PyPI projects were compromised... The discussion of this behavior began during PEP 740 (Digital Attestations) back in January 2024. The discussion was restarted in March 2026 after the popular packages LiteLLM and Telnyx were compromised. These packages were compromised due to a "mutable reference" in these projects' usage of the Trivy GitHub Action...

To quantify how disruptive this change would be to existing workflows, the PyPI database was queried for projects that have published new files to old releases... [O]nly 56 projects of 15,000 had published a [Python] 3.14-compatible wheel more than 14 days after a release was available. This topic was brought to the Packaging Summit at PyCon US 2026 by PyPI Safety & Security Engineer, Mike Fiedler. The rough consensus of the discussion was that the summit attendees thought it was "acceptable to require users to bump to the next version" to support new Python versions. With the data and consensus in hand, Seth moved forward with a patch to reject new files on old releases which was merged July 8th, 2026.

An anonymous reader quotes a report from Ars Technica: We are a month away from the initial rollout of Google's Android developer verification system, and the company contends this policy does not impinge on the platform's open nature. Still, the restrictions will be a big change, and there are still some unanswered questions. An issue that has come up repeatedly in the run-up to verification is what will happen to devs who can't verify because of where they live. It turns out that Google has a cryptic answer for that buried in an FAQ. Developer verification will soon block the installation of apps from unverified developers on any Android device running Google services, which is functionally all Android phones outside Russia and China. Developers who want to keep releasing software, even if it's not in the Play Store, have to provide Google with their ID and pay a small fee.

But what if you're an Android developer living in a sanctioned nation? Currently, the U.S. sanction list includes Iran, Cuba, North Korea, and occupied areas of Ukraine. Given the current uncertain state of US foreign policy, that list could change in the future. Google doing any business with developers in those places is a thorny issue, and it seems like the company has decided to just leave them hanging. A rather lengthy FAQ a few levels deep on the Google developer site addresses various issues around dev verification. Smack in the middle is this: "How does this program impact developers in sanctioned countries? Devices in sanctioned countries will be excluded from Android developer verification checks. This allows any developer to continue distributing apps in these regions without verification, though users there won't benefit from the enhanced security benefits of the program."

[...] A Google spokesperson has expanded on the FAQ and confirmed to Ars that people living in sanctioned nations will not be allowed to go through the verification process. That means they will not be able to effectively distribute software through any channel internationally. Today, someone making an app in, say, Cuba can distribute it freely around the world, as well as at home. Anyone can install it and see their work in action after tapping through a few sideloading alerts. In the coming months, that will no longer be the case. These unverified apps will only be easily installable in the sanctioned countries where verification doesn't exist.

An anonymous reader quotes a report from Reuters: OpenAI has discovered other instances in which autonomous agents have escaped containment as the company expands its investigation of the hacking incident at tech firm Hugging Face that drew global attention this month, two people familiar with the matter said on Friday. The new breakouts were uncovered during the company's publicly announced investigation into how one of its agents escaped what was meant to be a contained testing environment this month, the two people said, and OpenAI is now looking into those instances as well. One of the sources said that the escapes were limited in nature and that none of the agents were thought to have left OpenAI's network.

An OpenAI spokesperson referred to a statement issued by the company on Tuesday that said it was reviewing "broader activity from our models" in addition to the Hugging Face intrusion. The discovery of additional rogue behavior at OpenAI, even if limited in nature, could feed growing appetite for regulation coming out of the White House and elsewhere. The expanded investigation by OpenAI was launched shortly before its primary rival, Anthropic, disclosed that its models were also responsible for a series of break-ins that led to breaches at three other companies dating back to April, according to the two sources and a third source familiar with the matter. The recent discovery of other past breakouts at OpenAI has not previously been reported.

AI safety experts said the new disclosures paint a portrait of a group of cutting-edge labs whose ability to develop dangerous autonomous hacking agents outstrips their ability to keep them under control. "We have a whole industry where the people designing, developing and putting out these tools aren't keeping up themselves to responsibly develop these things and keep them safe," said Maurice Chiodo, a mathematician who works at Cambridge University's Center for the Study of Existential Risk. Reuters could not establish exactly how many incidents OpenAI investigators found or the timings or circumstances under which they occurred. The three sources said OpenAI and outside experts were examining log data from earlier in the year in a bid to understand what took place.

An anonymous reader quotes a report from NBC News: Cyberattacks targeting municipal water systems have been reported in at least seven states this week, prompting the FBI and the Environmental Protection Agency to warn utilities nationwide that hackers are trying to disrupt critical water infrastructure. In a public service announcement Thursday, the agencies said water and wastewater utilities have reported incidents to the FBI, with some malicious activity degrading water operations. The announcement does not name the states.

The warning comes after hackers targeted more than 30 municipal water facilities in Minnesota in an attack that had hallmarks of Iranian meddling, according to a law enforcement official. It is still under investigation. A spokesperson for Minnesota's information technology services agency said Thursday there was no indication the breaches contaminated any municipal water supplies. The federal Cybersecurity and Infrastructure Security Agency said in a separate alert that some larger attacks on water infrastructure had "resulted in boil water notices and sustained manual operations," though it did not say where.

[...] The federal advisory said the malicious cyber actors, or MCAs, targeted specific brands of control systems used by municipal water utilities, though the FBI and the EPA urged operators of all systems to take precautions. [...] The agencies said the hackers remotely accessed internet-facing devices, changed IP addresses and passwords, and caused utilities to lose monitoring and control capabilities. The federal advisory calls on system operators to remove programmable logical controllers, or PLCs, from direct internet exposure by putting them behind secure gateways and firewalls; use strong passwords; and limit communications between authorized control system devices through access control lists.

Google says AI-assisted workflows helped Chrome fix 1,072 security bugs across versions 149 and 150, more than the previous 23 releases combined. The company is also testing twice-weekly security updates and "dynamic patching," which could apply most fixes without requiring users to restart the entire browser. "By leveraging Chrome's multi-process architecture, dynamic patching sequentially replaces background child processes (like the Renderer and GPU) with updated binaries on the fly," says Google in a blog post. PiunikaWeb reports: Alongside dynamic patching, Google is rolling out smarter background updates during periods of minimal user disruption. Starting with Chrome 150 on macOS, the browser takes advantage of the operating system's windowless state. If all Chrome windows are closed but the app remains running in the background, the browser will quietly auto-restart to apply pending updates.

For enterprise environments, IT admins can continue to manage fleet-wide deployments through Chrome Enterprise Core or enforce update prompts using the RelaunchNotification policy. Google's long-term vision is a browser that remains continuously protected in the background without interrupting your daily browsing session. In the meantime, you can manually trigger pending updates by clicking the update prompt in the top-right corner of Chrome.

A production company and filmmaker are suing Netflix for $105 million, alleging the streamer lost a stolen drive containing an unencrypted master copy of the unreleased Nicolas Cage film Fortitude, which they claim damages its exclusivity and market value. Netflix denied responsibility for the lost film but said it takes content security seriously and has offered to monitor piracy sites for unauthorized copies. CBS News reports: The complaint filed on Wednesday in California district court alleges that the film's associate producer, Daniel Haido, hand-delivered an unencrypted master copy of the film to Netflix so the company could screen it as a potential buyer. Haido verbally instructed the employee to delete the files after the screening, according to the suit. A little over a week after the screening, Netflix emailed the filmmakers to say the drive had been stolen, the plaintiffs allege. "Someone stole a good amount of drives from our office desks this past week," a Netflix executive wrote in the email, according to the suit.

The complaint notes the movie, entitled "Fortitude," took over seven years to make and cost $45 million. It tells the story of a secret mission called Operation Fortitude during World War II that was orchestrated to mislead the Nazis about the Allied invasion of Europe. The film stars Nicolas Cage as Dusko Popov, a real-life spy during World War II, as well as Sir Ben Kingsley and Ron Perlman.

The plaintiffs said studios will now be dissuaded from buying the rights to the movie, knowing that a version of it could be released by a third party for free. "The film's value depended in significant part on its exclusivity as an unreleased, first-to-market work," the complaint states. "By losing control of the film, Netflix destroyed that exclusivity and materially, if not completely, impaired the film's marketability."

Anthropic found that Claude models breached three outside organizations during cybersecurity tests because misconfigured environments accidentally gave them access to the internet. The company notified those affected and urged other AI labs to audit their own testing systems. The BBC reports: Anthropic said in a statement that it reviewed more than 140,000 tests to find evidence that Claude - its family of AI models - could access the internet from testing environments that were designed to be sealed off. The tests include so-called "capture-the-flag" evaluations in which Claude was tasked with obtaining information by breaching other systems - a common way that experts assess a model's hacking capabilities.

A "misconfiguration" on systems run by Anthropic and its testing partner left the models with live internet access, allowing them to breach other systems, the San Francisco-based firm said. Anthropic said the earliest incidents date back to April and that it is "approaching the fixes as if the responsibility were ours alone." Neither Anthropic nor the organizations that were breached had noticed the intrusions at the time.

Anthropic said it could have reviewed its records more thoroughly and added that the findings gave the firm "cautious optimism" that such risks can be overcome with more investment and tighter measures. "The broader lesson is not necessarily that AI has developed a fundamentally new attack capability," cyber security expert David Allott told the BBC. "Instead, it is that AI agents can combine capabilities, obtain credentials and system access to take actions autonomously, while adapting scope and scale at machine speed," he added. The announcement comes just days after OpenAI said that its models had breached the systems of other companies, including AI tools platform Hugging Face.