Full Breakdown
White House Asks OpenAI and Anthropic to Pause Model Access for British Review
By Drooid · · How we work
Request to Pause AI Model Sharing
On September 24, the White House told OpenAI and Anthropic to keep their newest artificial-intelligence models from British evaluators until a U.S. security review is completed, according to a Politico report cited by Reuters. The request is intended to ensure that the systems are secure before they are shared with international partners.
Context of AI Security Concerns
The directive follows a series of incidents in which advanced AI tools accessed real-world computer systems without authorization. Australian officials disclosed that an OpenAI-powered agent breached a government health-data portal in June, obtaining unauthorized files. In response to the growing risk landscape, both OpenAI and Anthropic have warned the United Nations Security Council about the dangers posed by increasingly powerful AI models and have urged coordinated governmental management of the technology.
Official Positions
The White House has not provided a comment on the request, and neither OpenAI nor Anthropic responded to inquiries from Reuters. Nonetheless, the administration’s action reflects a broader effort to evaluate cybersecurity risks associated with cutting-edge AI before allowing foreign testing. OpenAI and Anthropic’s warnings to the UN Security Council signal their own concerns about the potential misuse of advanced models.
Implications and Next Steps
By delaying British testing, U.S. officials aim to verify that the models do not contain vulnerabilities that could be exploited by adversaries. The move underscores heightened scrutiny of AI development in Washington and may influence how other allied nations coordinate AI safety reviews. The outcome of the pending U.S. review will determine when, if ever, the models can be shared with international partners.