Sharing of New Models with UK AI Safety Institute Suspended
"U.S. Government to Review American Companies' Models First"
Clash with Previous Practice of Pre-Release Safety Verification

The White House has requested that AI companies Anthropic and OpenAI submit any new AI models for review by the U.S. government before providing them to a British research institute, Politico reported on September 24 (local time).

Anthropic logo. Reuters Yonhap News

Anthropic logo. Reuters Yonhap News

View original image

According to the report, the White House recently asked the two companies not to share any new AI models with the UK’s AI Safety Institute (AISI) until the U.S. government has completed its own testing.


A senior official from the Trump administration told Politico, “The United States wants to review new models and ensure their safety before they are shared with our partners. These are American companies.”


Established in 2023, the UK’s AI Safety Institute has been working with major AI firms to conduct safety tests on new models before they are released to the public. Due to this request, companies now face increased pressure to decide whether to proceed with the UK’s pre-verification as usual or to follow the U.S. government’s policy.


In fact, it is known that Anthropic recently chose not to provide its new AI model, Claude Mythos 5.1, to the UK side, releasing it only to select American institutions. Henry de Zoete, Director of the UK’s AI Safety Institute, also stated in a letter submitted to the British Parliament this month that the institute has not secured access to the model.



The UK government maintains that cooperation with major AI companies is ongoing. Ed Miliband, UK Secretary of State for Foreign, Commonwealth and Development Affairs, emphasized at the United Nations Security Council on September 23 that governments must ensure sufficient transparency so that advanced AI models can be rigorously vetted and corporate activities properly assessed.


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing