US Seeks AI Security Review
White House Requests AI Firms Withhold New Models From UK for Security Review
Anonymous
The White House has reportedly asked OpenAI and Anthropic to withhold new AI models from the UK’s AI Security Institute until U.S. officials review them for cybersecurity risks. Anthropic has so far withheld Claude Mythos 5.1, while the institute has previously tested some frontier models before release, including an OpenAI model. The request comes amid concern that advanced AI systems can enable cyberattacks; Australian officials recently said an OpenAI agent accessed files through a government health data portal. The move could disrupt the UK’s established access to pre-release models and reflects growing treatment of frontier AI as a national-security concern, though the companies and White House had not publicly commented in the reports.
Key Points
1
A senior U.S. administration official said the review request reflected a policy applied to each new frontier model: “Because they’re American companies and this has been our policy with every new frontier model that comes out.”
2
OpenAI and Anthropic had also warned the UN Security Council about risks from increasingly powerful AI systems and urged governments to cooperate in managing the technology.
3
A separate report says a U.S. Commerce Department export-control directive in June 2026 targeted Anthropic’s Fable 5 and Mythos 5 models, citing the risk that jailbreaks could compromise safety protocols; Anthropic then suspended foreign access to those models.
4
The same report says OpenAI restricted its GPT-5.6 Sol model to customers approved by the U.S. administration.
Sentiment
Mostly PositiveNEGATIVENEUTRALPOSITIVE
Publishers
28
Articles
110
Reach
138
A
AnonymousAuthor
