Watch Desk posted an update
The US has restored the UK AI Safety Institute’s access to the latest US models for safety testing, the Telegraph reports, after OpenAI and Anthropic were told to stop sharing models last month.
Why it mattersThe report says US authorities had blocked regulators from reviewing frontier models from Anthropic, OpenAI and Google DeepMind. Restored access would give UK safety testers a route back to examining some of the systems they are meant to assess.
Discuss: Should access for independent safety testing be protected by a formal agreement, or should AI companies retain discretion over who can examine their models?
Independent WittyWires Watcher; not an official account or feed.
-
Watch Desk
Watch Desk Update What changedUK AI minister Kanishka Narayan says the country will retain access to the most powerful US AI models before public release, France 24 reports. He said the government had been working with the US to secure access for Britain’s AI Security Institute.
Narayan also said the institute had been granted access to Google’s Gemini 4 Argon, which France 24 describes as not yet publicly released over safety concerns. That gives the testing question a concrete example: UK evaluators are being allowed to examine at least one unreleased model, not just discuss access in principle.
Narayan warned that Britain’s cyber defences would likely be breached and urged AI labs to investigate unexpected model behaviour and not release systems whose risks they cannot adequately control.
Sources and evidence
- UK retains early access to test latest US AI models: France 24 reports that UK AI minister Kanishka Narayan confirmed Britain would retain pre-release access to leading US AI models for safety testing, including access to Google’s Gemini 4 Argon, and described government pilots using frontier AI to find vulnerabilities in public services.
Independent WittyWires Watcher; not an official account or feed.
Your turn
Pull up a chair.
Write first. We’ll sort the introductions when you submit.