The US has reversed its decision to block British officials from accessing unreleased AI models, restoring the UK’s ability to test systems from Anthropic, OpenAI and Google DeepMind before their public release, according to a report published by The Telegraph.

The reported reversal follows lobbying by Kanishka Narayan, the UK’s AI minister, after the White House asked American developers last month to stop sharing their latest models with the UK’s AI Security Institute (AISI). The restriction had prevented British officials from reviewing Anthropic’s Claude Mythos 5.1 model.

The AISI was established following the first international AI Safety Summit in 2023 and evaluates advanced AI systems for their capabilities and potential misuse. Its access to unreleased models allows officials to assess risks before the systems become publicly available.

Narayan is expected to announce new measures to strengthen the UK’s response to AI-related security threats in a speech on Monday. Proposed changes to the Cyber Security and Resilience Bill would strengthen reporting requirements for companies involved in critical national infrastructure when security breaches involving AI occur.

The government will also increase its use of advanced AI systems to protect government departments against cyber attacks, while providing additional resources for cybersecurity officials in the National Cyber Security Centre, Cabinet Office and Department for Culture, Media and Sport to investigate and respond to incidents.

The announcements follow warnings from the heads of MI5, MI6 and GCHQ, who reportedly told ministers that AI was giving hostile actors and terrorists new capabilities to target sensitive sites and obtain confidential information.

The restoration of access comes as the UK seeks to establish itself as an international leader in AI governance. Speaking at the United Nations General Assembly last month, Andy Burnham said Britain would use its presidency of the G20 in 2027 to encourage international cooperation on AI standards.

Burnham said the UK wanted to act as an “honest broker” in developing “a single set of global principles and standards” for AI models. The earlier US restrictions had complicated that ambition by limiting British officials’ ability to examine some of the most advanced systems developed by American companies.

US President Donald Trump has previously opposed international restrictions on AI development. Addressing the United Nations last month, he said he wanted to “encourage” AI, “not rein it in”.

The reversal follows signs of a more nuanced US approach to AI security. Reuters reported on 11 October that Trump’s newly launched AI task force had said developers must report security incidents involving their models, although the White House and the companies involved had not immediately responded to requests for comment on the reported change in access arrangements.

---