Key Points:
- Downing Street has officially rejected reports claiming Prime Minister Keir Starmer petitioned Washington for an exemption from the latest US artificial intelligence export restrictions.
- The sudden US export directive forced AI developer Anthropic to suspend global access to its advanced Claude 5 family models, Fable 5, and Mythos 5.
- The regulatory action stems from concerns that users could bypass safety protocols to exploit powerful cybersecurity capabilities within the restricted software.
- The cutoff has sparked debate over Britain’s reliance on American technology giants and its struggle to establish sovereign computing power.
The British government has officially denied reports that Prime Minister Keir Starmer attempted to secure a special carve-out from sweeping US export controls on artificial intelligence. Speculation had mounted that London was actively lobbying Washington for an exemption to maintain access to Anthropic’s most advanced software systems. However, a spokesperson from Downing Street declared those reports categorically untrue. The swift denial comes as G7 leaders gather in France to address growing anxieties about technological sovereignty and the sudden restrictions placed on critical computational tools.
The controversy began when the US Commerce Department’s Bureau of Industry and Security issued an unexpected export directive. This regulatory action forced Anthropic to disable global access to Fable 5 and Mythos 5, two of the most capable models in its Claude 5 family. Rather than risk non-compliance, the San Francisco-based AI firm took the drastic step of pulling both models from service entirely. The shutdown cut off not only international customers but also affected foreign nationals working within the US, exposing the sheer reach of American export laws under the 2018 Export Control Reform Act.
US officials restricted the exports after raising major cybersecurity alarms. Specifically, authorities feared that users could “jailbreak” Fable 5—which had a wider public release with safety safeguards—to access the far more potent capabilities of Mythos 5. Mythos 5 was designed with highly specialized features capable of scanning computer networks to identify hidden software vulnerabilities. While developers intended for this to help organizations patch and defend their networks, defense analysts warned that bad actors could exploit the same tools to launch sophisticated cyberattacks against financial institutions and national infrastructure.
Anthropic executives have strongly disputed the US government’s assessment of the security risks. The company argues that the evidence presented by Washington concerns only a very narrow, fixable vulnerability rather than a fundamental flaw in the models’ architecture. They warned that applying such an aggressive regulatory standard could halt the public release of frontier AI systems altogether. Despite emergency meetings in Washington between company leaders and officials from the US Commerce Department, the restrictions remained firmly in place, leaving the industry divided over how to balance public safety with rapid innovation.
Before the official denial from Downing Street, reports indicated that Starmer’s chief AI adviser, Jade Leung, was spearheading a quiet effort to get Britain included on an approved list of nations. The goal was to secure a targeted, surgical approach to replace the blanket export ban. The situation put the UK government in an awkward diplomatic position. While the Prime Minister’s team maintains that they did not formally request an exemption, the episode has clearly highlighted the immense dependence of British businesses—particularly small and medium-sized enterprises—on software controlled by a tiny handful of American corporations.
This technological divide quickly became a focal point at the G7 summit in France. During a working lunch with technology executives, European Commission President Ursula von der Leyen urged the US to share access to frontier AI models. She emphasized that a trusted partnership is in the mutual interest of both Western economies. However, US officials have remained firm, with some administration members stating that granting exceptions to individual G7 allies is illogical and would undermine national security strategies.
For many industry observers in London, this sudden disruption has exposed a massive sovereignty gap. The UK currently lacks its own frontier model developers, leaving its tech sector highly vulnerable to policy shifts in Washington. To combat this vulnerability, the UK government previously committed £1.1 billion to back domestic semiconductor design and supercomputing infrastructure. Additionally, plans exist for a £500 million Sovereign AI fund. However, these investments look very modest when compared to the hundreds of billions of dollars that US technology giants spend on building and training AI models every year.
The tightening of centralized US export controls has also triggered a notable shift in the broader technology market. Almost immediately after Anthropic pulled its models, decentralized AI infrastructure tokens began to rally. Alternative decentralized networks saw surges in value ranging from 14% to 21%. Many smaller tech firms and independent developers are now actively exploring these decentralized alternatives to build products that cannot be disabled by a single regulatory pen stroke.
As the diplomatic dust settles, the immediate future of advanced AI access for UK enterprises remains highly uncertain. Without a formal path to access restricted US models, British tech startups must either work with older, less capable software or wait for Anthropic to resolve its regulatory dispute with the US Commerce Department. The episode serves as a stark reminder of the fragile foundation upon which much of the modern digital economy is built. For Keir Starmer’s administration, the challenge will be finding a balance between nurturing a domestic tech ecosystem and managing relations with a highly protectionist US government.
