The U.S. reversed a policy that required White House approval before AI developers like OpenAI and Anthropic could share new models with Britain's AI Security Institute. UK AI Minister Kanishka Narayan announced on October 12, 2026, that Britain worked closely with Washington to secure pre-release access for safety testing. The move signals a softer stance by President Donald Trump's administration on international AI oversight.
US policy reversal on pre-release testing
- ▪American AI developers including OpenAI and Anthropic are permitted to provide the British government early access to artificial intelligence models for safety testing without violating U.S. policy
- ▪A prior U.S. policy required White House approval before leading artificial intelligence models could be shared for pre-release safety evaluation
- ▪The Trump administration reversed a prior policy that requested American AI developers not to share new artificial intelligence models with Britain's AI Security Institute prior to public release
UK announcement of model access agreement
- ▪UK AI Minister Kanishka Narayan announced on October 11, 2026, that Britain worked closely with the U.S. government to secure pre-release model access for Britain's AI Security Institute
- ▪The United Kingdom will evaluate leading artificial intelligence models prior to their public release under a bilateral government understanding announced on October 11, 2026
Trump's stance on AI safety
- ▪U.S. President Donald Trump's AI task force announced on October 9, 2026, that artificial intelligence developers must report security incidents involving their models
- ▪U.S. President Donald Trump has repeatedly dismissed calls for binding safety requirements on artificial intelligence developers
Debatable claims
- ▪Pre-release safety evaluation by independent institutes is an effective way to mitigate AI risks
- ▪US AI developers should share pre-release models with Britain's AI Security Institute
Story comments
Loading comments…