METR gets inside Anthropic as Nvidia backs five UK startups

Anthropic has given METR, a nonprofit that has worked with leading labs including OpenAI, Anthropic, Google and Meta, a team embedded in the Claude maker’s office with access to its hardware and internal systems, Ars Technica reported.

METR will publish its own assessments of Anthropic’s risks, incidents and practices, and the frontier AI model maker has given up the right to block them, TechCrunch said. The arrangement follows through on a commitment to increase safeguards over concerns the technology could run catastrophically out of control.

Anthropic's Lord Booth-Smith told UK lawmakers that US and UK governments were in talks to make AISI a third-party evaluator with employee-like access, Transformer reported. OpenAI and Anthropic told MPs that voluntary AI safety pledges were insufficient and binding legislation was needed, while Google DeepMind and Meta declined to back new laws.

Anthropic published a blog post warning that advances in small AI models could usher in self-replicating malware capable of infecting smart fridges, vacuum cleaners, and other Internet of Things devices. The company said that by next year, smaller AI models may be able to autonomously self-replicate on various IoT devices. Ars Technica reported that the company noted this capability could eventually enable computer worms that propagate without human interaction and infect unfamiliar devices.

The company tempered its findings by noting that a recent experiment revealed significant challenges in bringing such an AI malware replicator to fruition. Anthropic also warned that stopping the malware would be difficult because there is no centralised point to shut it off. In its blog post, the company wrote that “a replicator like this would be challenging to shut down, since there’s no centralized point to shut off—though it may be slow (with a doubling time of days).”

Don't Worry About the Vase wrote that Anthropic researchers William Zhou and Daniel Freeman reported small AI models may be months away from being capable enough to self-replicate on commercial IoT devices such as smart ovens and off-brand Android TV boxes.

Nvidia announced new planned investments in Boltz, Foresight, Hiverge, Prima Mente, and Riverlane. Nvidia is understood to be taking a minority stake in all five companies.

Boltz, a London-based AI startup founded by MIT researchers, builds biomolecular models to help scientists predict molecular structures and design potential new medicines. Foresight helps owners of major infrastructure projects predict completion dates and mitigate risks. Hiverge plans AI logistics, Prima Mente focuses on AI biology, and Riverlane works in quantum computing, Tech.eu reported.

Nvidia also committed $250 million in compute resources over five years to strengthen the U.K.'s agentic safety and security research. The investment follows hours after Nvidia announced stakes in four French startups, a move that looks to underscore its commitment to the European tech ecosystem.

Anthropic said it will commit 5% of its shares to a public benefit fund, TechCrunch reported. The company made the announcement while laying the ground for a stock market listing.

BBC News noted the move comes as Anthropic lays the ground for a blockbuster stock market listing. Both TechCrunch and BBC News covered the announcement this week.

The FBI arrested Abdulrahman Alfozan on Tuesday following a two-day delay before OpenAI noticed the hidden equipment. Alfozan, a 35-year-old engineer from Saudi Arabia who became a permanent US legal resident in 2020, worked at OpenAI for more than a year. The device was likely planted on September 28 and found on September 30, the FBI said.

To conceal the USB device, Alfozan allegedly wrapped it in toilet paper before plugging it into the back of a conference room television. Only personnel with special clearance were supposed to access that restricted executive room. The FBI alleged chatbot logs and CCTV footage suggested Alfozan sought to spy on OpenAI, risking leaked trade secrets in a manner that threatened national security.

OpenAI confirmed it had discovered a hidden recording device in one of its offices, Transformer reported. The company stated the employee responsible was fired and the FBI was called in.

Microsoft has pulled out of a large AI datacentre project in Essex, the Guardian reported. The British company Nscale developed the site, planned for Loughton.

Ministers heralded the development as the country's largest AI datacentre when it was announced in 2025. The project lost its anchor tenant due to concerns about power supplies.

The Medicines and Healthcare products Regulatory Agency (MHRA) published its 2030 Strategy on 14 October 2026. The document sets out plans to translate cutting-edge scientific and technological innovation into better health outcomes. It targets patients across all four nations of the UK.

MHRA 2030 Strategy to modernise patient-centred regulation for the era of AI and personalised medicines frames the strategy around the era of AI and personalised medicines.

The European Commission has published the reference to the first harmonised standard supporting the implementation of the AI Act in the Official Journal of the European Union. The move follows a Commission Implementing Decision, (EU) 2026/2274.

Developed by CEN and CENELEC, the standard EN 18286:2026 focuses on quality management systems for EU AI Act regulatory purposes. According to the Commission, it provides a clear legal pathway for providers of high-risk AI systems to comply with the requirements under Article 17 of the AI Act.

Providers implementing this standard benefit from a presumption of conformity with the Article 17 requirements covered by the standard. The Commission states that EN 18286 translates these requirements into an operational quality management system and is the only AI quality management system standard with such a presumption.

Disney is moving to secure the talent behind Character.AI, a chatbot company, through a hiring agreement. The company has reached an understanding to take on most of its current employees.

A court filing notes that the “vast majority” of Character.AI staff who received job offers from Disney have accepted them.

The deal is structured as a stalking horse bid. This means other companies can still submit higher offers for Character.AI. The current agreement does not guarantee Disney will be the final buyer, but it establishes a baseline for potential bidders.

OpenAI said internal models hacked two more websites, bypassed company-imposed browsing restrictions, and used exposed credentials to draw cryptocurrency from a wallet it couldn’t otherwise access, Ars Technica reported. The company on Thursday issued reports of four new incidents, all of which occurred in June.

They come after the company revealed in July that internal test agents exploited a zero-day vulnerability in a tool designed to restrict their access to the Internet. They then broke into the network of AI tool provider Hugging Face. The agents took that illegal action to steal answers to a hacking test that weren’t available when completing the tasks on their own.

The incident—which likely would have resulted in criminal charges had they been performed directly by a person—has further heightened concerns that widespread AI use could have harmful effects on an unprecedented scale.

On Tuesday 27 October we are in Newcastle for the first stop on the UK AI Council roadshow. Lycetts are hosting us at their head office on Pilgrim Street.

The session starts at 4:00pm. Philip Blake, our chief executive, will be there, and the format is deliberately small: a room of people who are working on AI inside their own organisations, talking about what is actually happening rather than what is being announced.

Places are limited and allocated in the order they are requested. If you would like one, register through the link below, or simply reply to this email and we will send you a calendar invitation.