The Register

OpenAI says it will deny internet access to the frontier models it is training, a move that underscores the inadequacy of safety measures undertaken to date."Based on the current state of alignment research, monitoring systems, and the security of internet infrastructure, we believe the responsible approach is to conduct frontier model training with no internet access," the company said in a blog post. "We've implemented this policy and will maintain it until advances in these safeguards justify

As the PHP2Shell chaos continues, researchers have found a third PHP vulnerability that could allow attackers to take over websites or remotely run malware on vulnerable systems. But this one won’t receive an emergency fix.The PHP Foundation on Friday told The Register that it has reserved a CVE identifier (CVE-2026-115560) for a third RCE flaw to go with the first two that comprise PHP2Shell, and it will release a full advisory and fixed software with the next regular PHP release.CVE-2026-11556

An independent evaluation of OpenAI's internal misalignment monitoring system revealed that the system was not fit for purpose during a period when the company's bots were breaking into model-hosting platform Hugging Face. The safeguards weren't nearly strong enough."The METR and Redwood employees discovered a number of novel vulnerabilities in the asynchronous internal misalignment monitoring system as it existed on and before July 21st," said evaluation partners METR and Redwood Research in a

If you've ever wondered what the bots think of us puny humans, you'll be disappointed in the answer. An AI ranking outfit asked 12 different LLMs about people, and every single one agreed: We’re not as rational and intelligent as we think. “A person is smart,” Tommy Lee Jones’ Kay opines in the 1997 movie Men in Black. “People are dumb, panicky dangerous animals,” he goes on to argue, and that seems to be what a dozen models from ten AI labs in three countries also believe about humanity, accord

Most programming languages were written for human programmers, but that's about to change. University of Michigan professor Jason Mars has developed an academically ambitious language called Jac, and a runtime called Jaseci, that treats AI and AI agents as first-class entities. Mars demonstrated the language and runtime in a talk at the All Things Open conference, held in Raleigh, North Carolina, this week.The tools we use to create software – everything from programming languages to build toolc

The Arc browser has flatlined, with its maker announcing it will be sunsetting it altogether on December 22, 2026. The Browser Company announced the news on October 22, just over three years after we noted Arc's arrival. Nothing will happen immediately, other than an end to new Arc accounts. Starting November 23, lights will begin turning off on iOS and Android, with several AI features disabled and sign-ins to sync tabs removed. Finally, on December 22, it will no longer be possible to access

SparkyOS is a novel distro from the folks behind SparkyLinux: it's a version of Arch that can you can use without installing it at all.SparkyOS is a new distro from the team behind the Debian-based SparkyLinux, but it's quite different. SparkyOS is based on Arch, and it runs entirely from USB. It's not a conventional live medium – this is a full local install, but shipped as an image of a USB key that you can boot and run. It comes with the Xfce desktop and an array of apps. The idea is that you

Microsoft ended support for Publisher this month, but users with orphaned documents still have options.Support for perpetual versions ended on October 13, but those copies will keep working without further security updates, bug fixes, or technical support. Publisher left Microsoft 365 subscriptions on October 1, although existing installations may remain accessible until removal updates arrive. Those start reaching Current Channel devices in mid-November and Monthly Enterprise Channel devices in

NASA's safety watchdog, the Aerospace Safety Advisory Panel (ASAP), had good news and bad news for the space agency, and SpaceX boss Elon Musk quickly disagreed with some of it.Musk disputed the reported suggestion that approximately 22 launches might be needed to support a landing by the Human Landing System (HLS) variant of Starship on the Moon. "That's way too high," Musk responded on X. "Actual number will be more like 6 to 8."The US lunar landers being developed for Artemis require refuelin

Linus Torvalds is embracing AI to write Linux kernel code, and some people are having fits. He has a message for them. So do I.Roy Schestowitz, editor of Tux Machines, summed up his reaction to the news as "Linus Torvalds Turns to Slop." In the podcast Pivot to AI, podcaster David Gerard agreed: "Linux kernel goes full AI slop code." Meanwhile, in the Tom's Hardware comments section, commenter "LordVile" wrote: "This is a publicity stunt from his AI overlords [The Linux Foundation] who pay his w

Working in the same place for what feels like millennia can have its benefits, however working for the same place for what seems like millennia carrying over all the benefits of your original contract can be even better than that.Some of those benefits are, admittedly, useless to me. I'm unlikely to avail myself of the Company-reimbursed landline (that all on-call personnel are entitled to), and I won't be needing the 2400bps modem or the VT100 for after hours support. The one month pregnancy le

Britain is threatening subsea cable saboteurs with up to 14 years in prison and unlimited fines, having apparently decided that Victorian-era laws aren't quite up to the job of keeping Russian submarines away from critical infrastructure.The government on Thursday set out plans to overhaul legislation dating back to 1885 and introduce tougher penalties for damaging the underwater cables and pipelines that keep Britain's internet and energy supplies running.According to Whitehall, subsea cables c

Today's victim of the bork fairy is a claw machine that even Android seems to have rejected for its alleged unfairness. And Google, as we all know, is all about fairness.Eagle-eyed Register reader Scott spotted the stricken machine in an arcade adjacent to a roller rink, a place our reader visited with a friend and described as "something my currently bruised body reminds me I hadn't done for about 20 years."We assume they were talking about roller-skating, and not the exertions of accompanying

The Agentic AI Foundation (AAIF) has been entrusted with stewardship of four additional AI projects in the hope of broadening the set of vendor-neutral tools for making AI agents.The AAIF was founded in December 2025 under the auspices of the Linux Foundation as a vendor-neutral organization for the benefit of AI companies more interested in riding a tide that lifts all boats than making waves.It launched with three projects under its supervision: Model Context Protocol (MCP), goose, and AGENTS.

Energy can be hard to find on Friday mornings, which is why The Register uses them to offer a fresh instalment of On Call – the reader-contributed column that shares your hopefully empowering tech support stories.This week, meet a reader we’ll Regomize as “Lachlan,” who told us he once did tech support for a senior exec who had a side hustle writing essays for Japanese university students attending UK universities.The exec asked Lachlan for help diagnosing a sudden inability to send or receive e

Researchers have developed a new exploit that shows how even Address Space Layout Randomization (ASLR) on an x86-64 server cannot stop the PHP2Shell vulnerability. VulnCheck, which made it available to only the company’s Initial Access Intelligence customers, also warned that threat actors are using AI to find even more ways to attack. The team told us that it’s a “significant advance” because up until now, ASLR has been a major barrier to exploiting PHP2Shell on real-world servers. They don't

Researchers have administered AI-designed, bacteria-targeting viruses to a person in what they claim is a first, but don’t press them for results: They’re not ready to say how it went quite yet.Radical Numerics, a San Francisco-based AI lab developing AI models for bioscience, announced the treatment this week, claiming it marks the first time a cocktail containing AI-designed bacteriophages has been administered to a human patient under a compassionate-use arrangement.These aren’t just any viru

Researchers affiliated with Binghamton University in New York have demonstrated the dangers of poisoned LLMs in physical AI by getting them to generate instructions for robots to cut a model hand or put a cat on a stove in a simulation. Fortunately, they also have a potential defense.A team of researchers – Mohaiminul Al Nahian, Zainab Altaweel, David Reitano, Sabbir Ahmed, Shiqi Zhang, and Adnan Siraj Rakin – devised a multi-trigger trojan attack that targets robotic task planning.The technique

GitLab on Wednesday patched a maximum-severity flaw that could allow unauthenticated attackers to execute arbitrary code on self-managed servers under certain conditions. Affected software releases date back a full 12 years - so upgrade to a fixed version now.The integer overflow bug, tracked as CVE-2026-110264, affects regular-expression compilation in GitLab Community Edition (CE) and Enterprise Edition (EE), including versions dating back to GitLab 7.0, released in June 2014.The flaw, “under

After embarrassing reports revealed weak private keys, one of which was just a single digit, Indian certification authority eMudhra has decided to take the L and get out of the global SSL/TLS cert game entirely.eMudhra announced Thursday that it was abandoning that arm of its business. Sure, the intense scrutiny heaped on the company was a factor, but the firm also argued that growing compliance costs and competition from free certificate providers had made the business increasingly difficult to