US reverses block; UK AISI regains pre-release access to frontier AI models
The US has reversed last month's order that told OpenAI, Anthropic and others to stop sharing new models with the UK's AI Security Institute, restoring the agency's pre-release access to frontier models from Anthropic, Google DeepMind and OpenAI, according to The Daily Telegraph and other outlets. UK AI Minister Kanishka Narayan said Britain had worked closely with the US to keep that access and announced a task force to protect UK government systems from AI-driven attacks, saying he wants Britain to be "one of the hardest countries in the world to attack using artificial intelligence." Reports say Anthropic had previously withheld its Claude Mythos 5.1 model from the institute; the White House, OpenAI and Anthropic did not immediately respond to requests for comment.
It shows a pre-release frontier-model safety testing channel can be restored after being cut for political reasons, though its durability still depends on the bilateral relationship.