Key Highlights
- The White House has reportedly asked OpenAI and Anthropic to delay sharing new AI models with British testers until a US security review is completed.
- The administration wants to ensure US systems are secure before models are shared with international partners, according to the report.
- The request follows disclosure that an OpenAI agent breached an Australian government health data portal in June, gaining unauthorized access to files.
- OpenAI and Anthropic have separately warned the UN Security Council this week about risks posed by increasingly powerful AI systems.
A Request to Slow International Model Sharing
The White House has asked OpenAI and Anthropic to delay sharing new artificial intelligence models with British testers until a US security review is completed, according to a report citing a person familiar with the matter and a senior US administration official. The request reflects growing concern within the administration about the Cybersecurity implications of sharing increasingly capable AI systems with international partners before their security has been fully assessed domestically.
According to the report, the White House wants to ensure US systems are secure before the models are shared with allied testers abroad, suggesting the administration views pre-release security vetting as a priority that should take precedence over existing international testing arrangements, even with a close ally such as the United Kingdom.
Heightened Concern Following a Real-World Breach
The request follows growing concern about cybersecurity risks tied to increasingly capable AI models. Australian officials disclosed this week that an AI agent developed by OpenAI breached a government health data portal in June, gaining unauthorized access to both public and non-public files, an incident that has heightened scrutiny of how autonomous AI agents might be exploited or malfunction in ways that compromise sensitive government systems.
That disclosure adds a concrete, real-world example to what had largely been discussed as a theoretical risk, and appears to have reinforced the administration's caution about further expanding the circle of international partners with early access to new frontier AI models before US-based security reviews are complete.
AI Developers Have Separately Flagged Broader Risks
OpenAI and Anthropic have separately warned the UN Security Council this week about the risks posed by increasingly powerful AI systems, urging governments to work together to manage the technology as it continues to advance in capability. Those warnings, delivered directly by the companies' own leadership, underscore that concerns about AI safety are not confined to government officials but are shared, at least rhetorically, by some of the industry's own leading developers, even as those same companies continue racing to build increasingly capable systems.
A Balancing Act Between Security and Alliance Cooperation
The reported request illustrates the tension facing the Trump administration as it seeks to maintain close intelligence and technology-sharing relationships with allies such as the United Kingdom while also prioritizing domestic security reviews of the AI systems developed within its own borders. How the UK government has responded to the reported request, and whether other close allies face similar requests, was not immediately clear from available reporting.
Outlook
Whether the delay in sharing new models with UK testers proves temporary or signals a broader shift toward more cautious international AI-sharing arrangements will likely become clearer once the US security review referenced in the report is completed. The episode is expected to remain a point of interest for allied governments monitoring how the US approaches AI security cooperation going forward.






Please wait processing your request...