On Oct. 12, 2026, the UK government said its AI Security Institute (AISI) was evaluating the latest generation of advanced AI models after work with leading developers and the U.S. government. The government announcement describes testing for government evaluation before public release. Quartz and Bloomberg reported a reversal of a U.S. approval requirement that had restricted sharing pre-release models with UK testers; Bloomberg also reported the reversal.
How UK access to pre-release models changed
Quartz and Bloomberg named OpenAI and Anthropic in their reporting on the restored access. DiarioBitcoin reported that the restriction was communicated on Sept. 24 and access returned on Oct. 11, after a 17-day interruption. Kanishka Narayan, the UK’s AI minister, credited lobbying and conversations between ministers and technology executives; Quartz reported that a summit with King Charles III partly facilitated those conversations.
The access is for government evaluation of models before public release. It is separate from an earlier U.S. dispute involving Anthropic’s Mythos models and federal AI safety guidance, covered in NeoTeo’s previous report.
Separate UK pilots addressed cyber risks
The UK government said two AI pilots identified and addressed more than 1,000 potential vulnerabilities across government systems and detected potential cyber threats. Those results came from separate cyber-defense work, not AISI’s evaluations of pre-release models. The government’s announcement describes the pilots alongside the evaluation program.
