White House orders OpenAI, Anthropic to withhold AI models from UK until US review
The White House has instructed OpenAI and Anthropic to undergo US government security reviews before providing their latest AI models to the UK's Artificial Intelligence Safety Institute (AISI) for testing. The request, issued by the Office of the National Cyber Director, aims to ensure US systems are secure before sharing with allies. Anthropic has reportedly paused providing its Claude Mythos 5.1 model to the UK AISI. The move comes amid US government debates on regulating frontier AI technologies.
IllustrationEditorial responsibility
- No named human review is recorded for this page.
- Reports are grouped by semantic similarity and deterministic rules. Language models may assist titles, summaries, translation and cross-source analysis; the page reads the event directly, while its address stays stable when the title changes.
- Summary covers the current reports
Cross-source coverage
Common ground
- Both sides agree that the US request to withhold AI models from UK testers is about control, not just safety.
- Both agree that the lack of a binding international framework for AI testing is a major problem.
- Both agree that the labs are making a commercial calculation by cooperating with US review.
- Both agree that neither the US nor the UK is villainous, but the system is broken.
Points of contention
- Neutral Agent says the US is being pragmatic to protect trade secrets and national security, while Western Agent says it's a power grab that undermines allied sovereignty.
- Neutral Agent argues the US request is about sequencing disclosure, not suppressing findings, but Western Agent says controlling the order is the same as suppression.
- Western Agent claims the US is preventing a binding international framework from emerging, while Neutral Agent says the US is acting rationally in a regulatory vacuum.
Blind spots
- Neither side fully addresses how developing nations are excluded from AI testing decisions that affect billions of people.
- Both overlook the chilling effect on global trust when allies see the US prioritizing its own review over shared oversight.
- The debate misses the need for harmonized legal protections for trade secrets across countries to enable fair, simultaneous testing.
WorldAttention’s read
The core issue is that the US is using its leverage over private AI labs to control the order of safety testing, prioritizing its own review over allied access. While the US has stronger legal protections for trade secrets, this creates a power imbalance that undermines trust and independent oversight. Both sides agree a binding international framework is needed, but they disagree on whether the US is blocking that or just acting pragmatically. The real blind spot is the exclusion of the global south and the lack of shared legal standards for commercial confidentiality. Until nations build a gate together, the gatekeeper will always be the one with the most power—and right now, that's Washington.
Reporting timeline
White House Requires US Review of New AI Models Before UK Testing Access
According to a Politico report cited by Gelonhui on September 25, the White House has instructed OpenAI and Anthropic to first undergo US government security reviews before providing their latest AI models to the UK's Artificial Intelligence Safety Institute (AISI) for testing. The requirement, proposed by the US Office of the National Cyber Director, aims to ensure US systems are secure before opening access to allies. Anthropic has reportedly paused providing its Claude Mythos 5.1 model to the UK AISI and is coordinating with the US government to gradually expand access to domestic and international partners. The UK AISI has historically worked with AI companies like OpenAI and Anthropic to conduct safety testing before public model releases. This development comes amid US government discussions on regulating rapidly advancing AI models, with some companies and experts warning of risks from insufficiently tested frontier models. UK Prime Minister Andy Burnham has called for international cooperation on AI safety standards, while the US is strengthening the Commerce Department's AI Standards and Innovation Center (CAISI) as a key AI safety governance body. OpenAI this week stated that CAISI should collaborate with international institutions including the UK AISI to advance global AI safety standards.
Read sourceWhite House Requires AI Models to Pass US Security Review Before UK Testing
According to a Politico report cited by Jin10, the White House has instructed OpenAI and Anthropic to undergo US government security reviews before providing their latest AI models to the UK's Artificial Intelligence Safety Institute (AISI) for testing. The requirement, proposed by the US Office of the National Cyber Director, aims to ensure that frontier AI models are assessed for US system safety before being shared with allies. Anthropic has reportedly withheld its Claude Mythos 5.1 model from the UK AISI and is coordinating with the US government to gradually expand access to domestic and international partners. The UK AISI had previously collaborated with AI companies like OpenAI and Anthropic, often conducting safety tests before public model releases. This development comes as the US government debates how to regulate rapidly advancing AI models, with some companies and experts warning of risks from insufficiently tested frontier models. UK Prime Minister Andy Burnham has called for international cooperation on AI safety standards, while the US is strengthening the Commerce Department's AI Standards and Innovation Center (CAISI) as a key AI governance body. OpenAI has stated that CAISI should collaborate with international institutions including the UK AISI to promote global AI safety standards.
Read sourceWhite House asks OpenAI, Anthropic to withhold AI models from UK testers until US review
Citing an anonymous US senior official and sources, a report says the White House has asked OpenAI and Anthropic not to share their latest AI models with the UK's AI Safety Institute until the US government completes its own testing. The request, issued by the Office of the National Cyber Director, puts the companies in a dilemma: either refuse the UK institute—one of the world's best-resourced government testing bodies that previously enjoyed priority access—or risk angering the Trump administration. The senior official stated that Washington wants to review and take measures to secure US systems before models are shared with allies.
Read sourceShow 2 older updatesHide older updates
White House asks OpenAI and Anthropic to pause new models from UK testers before US review
According to a report by Politico, cited by TradeAlpha, the White House has requested that AI companies OpenAI and Anthropic temporarily halt the release of new models from testers in the United Kingdom until a review is conducted in the United States. This action reflects ongoing U.S. government efforts to assert oversight over advanced AI development and testing, particularly concerning safety evaluations conducted abroad. The request underscores tensions between rapid AI deployment and regulatory review processes, as well as the international dimension of AI governance. The specific conditions and timeline of the pause were not detailed in the brief report.
Read sourceWhite House asks OpenAI, Anthropic to pause new models from UK testers before US review
According to a report by Politico, as cited by financial data platform Jin10, the White House has requested that AI companies OpenAI and Anthropic temporarily halt the release of new AI models from testers in the United Kingdom until those models undergo a review by U.S. authorities. The request indicates a proactive stance by the Biden administration to assert oversight over advanced AI development, particularly concerning models being tested abroad. The pause is intended to ensure that U.S. safety and regulatory standards are applied before any potential deployment or broader release. The specific models or timelines involved were not detailed in the brief report. This move reflects growing government scrutiny of frontier AI capabilities and cross-border testing practices.