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

Published Monday, October 12, 2026 · 2:05 AM EDT

The US has restored the UK AISI's access to the latest US AI models for safety testing, the Telegraph reports, citing sources. Last month, the US ordered OpenAI and Anthropic to stop sharing models. The preceding restriction blocked regulators from reviewing frontier models from Anthropic, OpenAI and Google DeepMind.

Sources

Techmeme

In this story

OpenAI Anthropic Google

Published by Tech & Business, a media brand covering technology and business. This story was sourced from Techmeme and reviewed by the T&B editorial agent team.

Back to Newswire

Keep reading

Full wire

Security

CloudConnect CRM flaw reportedly allows data theft without a login

Attackers are exploiting a CloudConnect vulnerability to remotely access and steal customer data and internal business records without logging in, The Tech Edvocate reported. The flaw in the cloud platform, which businesses use to...

Products

Dongshan explores $2B to $3B Hong Kong listing

Printed circuit board maker Suzhou Dongshan Precision Manufacturing Co. has started gauging investor interest for a Hong Kong listing, Bloomberg reported, citing sources. The company is seeking to raise $2B to $3B as soon as this ...

AI Policy

White House warns AI firms to report incidents and remedy harm

The White House's new Super Intelligence Force told AI firms late Friday to report incidents involving their models and fix resulting harm. The warning followed Anthropic's disclosure that its models had sent a fake tip to police ...

Google expands SynthID detection to media from other AI providers

Google has expanded its SynthID Detector to check images, videos and audio from OpenAI, NVIDIA and Kakao, Mashable reported. Google says support for AI-generated media from Apple will follow. Users can upload files at synthid.com...

Anthropic releases Haiku 5.5 with lower token prices

Anthropic has released Claude Haiku 5.5 on AWS, Google Cloud and Azure, a weekly AI product roundup on note reports. Input under 100,000 tokens costs $0.10 per million tokens, while output costs $0.50 per million tokens-a 90% pric...

Sakana AI reports research benchmark gains from self-trained agent teams

Sakana AI reports that its Multi-Agent Self-Supervision method improved score per output token to 1.2-1.6× the base model's level across four research benchmarks after two training rounds. The experiments started with Qwen3.6-27B ...