US restores UK AISI access to latest US AI models for safety testing

According to a Techmeme report from October 12, 2026, citing Telegraph, the United States has restored the UK AI Safety Institute's (AISI) access to the latest US models for safety testing. The report states that this decision followed an order last month directing OpenAI, Anthropic, and Google DeepMind to stop sharing frontier models with regulators. Telegraph reporter Tony Diver, citing sources, said the US had previously blocked regulators from reviewing frontier models from Anthropic, OpenAI, and Google DeepMind, and has now allowed the UK AISI to obtain those latest models again for safety testing. The report does not specify when access was restored, which model versions are involved, or why the US changed its approach. The verifiable sequence of events so far is that the US first told the three companies last month to stop sharing frontier models with regulators, and then restored the UK AISI's access to the latest US models for safety testing.