US Allows AI Developers to Share New Models with UK for Safety Testing

The Trump administration is allowing US artificial intelligence developers to grant the UK early access to new models for safety testing. The decision, published on Sunday, marks a shift in the policy of Donald Trump, who had previously opposed strict international oversight.

According to a report in The Telegraph, Washington has backed away from a previous requirement demanding companies like OpenAI and Anthropic not to share new models with the UK Artificial Intelligence Safety Institute prior to their official release. Companies can now share the information without fear of violating White House guidelines. Spokespeople for the president and the companies declined to comment on the reports.

The UK Institute was established following the 2023 Global Safety Summit to examine advanced models to assess their capabilities and prevent risks and misuse. The US move provides developers with greater latitude to collaborate with London as policymakers worldwide attempt to regulate rapidly advancing technology.

Donald Trump previously rejected calls to impose mandatory safety requirements on AI developers, but a shift in his approach is now emerging. On Friday, a new task force established by the president ruled that companies will be required to report security incidents in their models, a step reflecting growing alignment with international monitoring efforts.