The US government has asked prominent artificial intelligence companies, OpenAI and Anthropic, to present their new AI models for review by American authorities before making them available to the United Kingdom's AI Safety Institute (AISI). This directive, originating from the White House's Office of the National Cyber Director, places these tech firms in a delicate position, forcing them to choose between complying with the US request or risking strained relations with the American administration. Officials emphasize that this policy applies to all new frontier models developed by US companies, aiming to secure American systems first before broader international access.
This development unfolds amidst escalating concerns regarding the safety and potential risks associated with advanced AI, exemplified by a recent security breach involving an Australian government website during AI model testing. The UK's AISI, established in 2023, has historically enjoyed early access to and strong relationships with major AI labs, positioning itself as a global leader in AI safety research. However, the US insistence on a pre-emptive review challenges this established dynamic, revealing a growing contention over which nation will lead in shaping global AI governance and standards.
US Prioritizes Domestic AI Security Review
The White House has issued a clear directive to major artificial intelligence developers, including OpenAI and Anthropic, to submit their cutting-edge AI models for comprehensive review by US government agencies before offering them to the United Kingdom's AI Safety Institute. This strategic move is rooted in a US policy designed to ensure that American technological infrastructure and national security interests are safeguarded against potential risks posed by advanced AI systems. The request, reportedly from the Office of the National Cyber Director, underscores a proactive approach by the US to establish a primary assessment of these powerful new models, highlighting a commitment to domestic security and control over emerging AI capabilities. This prioritization aims to mitigate any unforeseen vulnerabilities or misuse before these technologies are introduced to international partners.
This demand creates a challenging dilemma for US-based AI companies, as they must navigate between maintaining robust international collaborations, particularly with the well-regarded UK AISI, and adhering to their home government's stipulations. A senior US administration official clarified that this policy is consistently applied to all new, advanced AI models originating from American companies, asserting that the US government seeks to perform its own thorough evaluations to secure its systems. This emphasis on a preliminary US review is critical for the White House, as it seeks to address a spectrum of concerns, ranging from data security to broader societal impacts, ensuring that the deployment of these technologies aligns with national interests and safety protocols.
International Friction in AI Governance
The White House's request to delay UK access to new AI models has introduced a new layer of complexity to international collaborations in artificial intelligence safety, particularly impacting the UK's previously established role as a key testing hub. Anthropic has, in response, limited access to its Claude Mythos 5.1 model primarily to US organizations, with plans for broader international release coordinated with the US government. This action signals a shift from the UK AISI's earlier privileged position, where it routinely conducted pre-release testing of models like OpenAI's GPT-6 Astra. Despite this, UK officials maintain that they continue to foster strong relationships with leading AI developers and have access to a wide array of advanced models, reinforcing their capability in AI safety assessments.
This situation highlights a nascent competition between the US and UK over leadership in AI governance and standard-setting, as both nations advocate for comprehensive safety protocols for frontier AI models. The US aims to bolster its own Center for AI Standards and Innovation (CAISI) as a central authority for developing international AI standards, working in conjunction with other national institutes. Conversely, the UK continues to champion global agreements and principles, stressing the importance of international cooperation for AI safety, especially given that AI risks transcend national boundaries. This evolving landscape necessitates a delicate balance between national security interests, commercial imperatives, and the shared global objective of safe and responsible AI development.
