Anthropic gives METR internal access, with disclosure limits
A new oversight arrangement, not a model safety certification.
Anthropic2 min readFrom Riyadh to the world.
Policy & safety
Governance, independent evaluation, safety and accountability.
4 storiesSearch news
UK–Canada AI security partnership: what changes?
The Canadian minister’s visit places the research partnership’s renewal in a wider diplomatic setting; technical outcomes still require measurement.
UK Government / Government of Canada3 min read
Britain prepares for AI and agent risks
The 12 October announcement combines defensive AI, model evaluation and incident preparedness; related official documents explain the practical requirements.
UK Government3 min read
UK accepts 44 healthcare AI recommendations and opens a new sandbox phase
Reforms emphasize oversight throughout use; implementation extends into 2027.
UK Government3 min read