Pulse

Political risk / Jul 12, 2026 / 4 min

The McMahon Act Just Reached AI

On July 7, the UK Science and Technology Committee warned Britain may be cut off from frontier AI "at the whim" of allies without a sovereignty strategy — the same fortnight Washington blacked out Anthropic's Fable for 18 days, Beijing weighed curbs on Qwen, and CIA chief John Ratcliffe compared frontier models to "digital nuclear weapons."

Thesis Westminster just admitted Britain's frontier stack runs on allied permission — and neither a £500 million startup fund nor another AISI red-team report substitutes for models London can keep when Washington or Beijing changes the guest list.

British MPs just admitted the UK's frontier AI stack runs on allied permission — not sovereign code — weeks after Washington and Beijing proved they can kill model access overnight, and months before a proposed nuclear-style superintelligence treaty finds a single signature.

What's new: On July 7, the House of Commons Science, Innovation and Technology Committee published Science diplomacy: Sovereignty, strategy, and the global race (HC 62). On July 9, chair Dame Chi Onwurah delivered the committee's statement on the floor of Parliament.

The report's blunt finding: the government has "no coherent strategic framework" for tech sovereignty and "may not be able to count on its allies" for access to critical technologies.

  • Onwurah: "There is a global race for sovereignty in technologies like AI, whether the government recognises it or not."
  • Onwurah: "The government needs a realistic plan to achieve sovereign capabilities in critical areas or risk having its access cut off at the whim of its partners."
  • MPs cite US restrictions on Anthropic's frontier models as the live example — not a theoretical risk.

Why the timing stings: The warning landed as both superpowers tightened the valves.

  • June 12, 5:21 p.m. ET: Commerce ordered Anthropic to suspend Fable 5 and Mythos 5 for all foreign nationals. Unable to nationality-gate in real time, Anthropic disabled both models globally within hours.
  • June 26: Commerce cleared Mythos 5 for roughly 100 vetted defender organizations before broader restoration.
  • June 30: Commerce lifted export controls on Fable after an 18-day blackout.
  • Joe Hancock (Mishcon de Reya): "A frontier AI capability, sold commercially and relied upon operationally, can be withdrawn from every non-US user by unilateral government action, at a few hours' notice, with no transition period."
  • July 8: Reuters reported Beijing is weighing overseas restrictions on Alibaba's Qwen, ByteDance's Doubao, and Z.ai's GLM — the mirror move after Washington's guest lists.
  • June: Beijing blocked Meta's $2 billion acquisition of agentic startup Manus.

What MPs want — and don't: The committee demands a defined sovereignty strategy, supply-chain mapping, and industry collaboration. It does not recommend Britain build its own ChatGPT-scale foundation models.

  • Labour has dropped the prior government's "own-collaborate-access" framework from official publications.
  • Science Secretary Liz Kendall has talked about "reducing over-dependencies" and launched a £500 million Sovereign AI Fund in April — equity bets on UK startups, not a frontier model.
  • Mishcon's Hancock: "No domestically available model currently competes with Fable 5 or Mythos 5 on capability, and that gap will not close quickly."

The nuclear frame is no longer metaphor: National-security officials are treating frontier AI like fissile material.

  • June 30, John Ratcliffe (CIA director, AWS Summit): "It would be... not misplaced to refer to their capabilities as akin to digital nuclear weapons."
  • June 30: AWS announced a $1 billion credit program for US intelligence agencies.
  • November 2025: MIRI's Technical Governance Team published an International Agreement to Prevent the Premature Creation of Artificial Superintelligence — modeled on the 1968 Nuclear Non-Proliferation Treaty. No country has signed.
  • The Independent drew the historical parallel to 1946's McMahon Act, when Truman ended nuclear cooperation with Britain — costing £150 million and forcing the UK to build its own bomb.

What we cannot verify:

  • Whether Beijing will actually restrict Qwen exports or is negotiating leverage.
  • How quickly UK firms could failover if Washington re-imposes Anthropic-style controls during a domestic dispute.
  • Whether Labour will publish the sovereignty strategy MPs demanded before the next export shock.

Convina's view: Westminster finally said aloud what June's Fable blackout proved in software: frontier AI is a permissioned utility with a kill switch, not a commodity you buy on a credit card. A £500 million VC fund seeds startups; it does not give Whitehall a model it controls when Washington rewrites the guest list or Beijing locks the weights. Until Britain defines what "sovereign" means beyond borrowed American APIs — and builds fallback capacity that survives an 18-day blackout — it is replaying 1946 with better press releases and worse lead times.

Research Signals

https://committees.parliament.uk/committee/135/science-innovation-and-technology-committee/news/214716/government-must-set-out-strategy-to-achieve-sovereign-ai-capabilities-uk-risks-being-cut-off-at-whim-mps-warn/ https://whatson.parliament.uk/event/cal56134 https://www.mishcon.com/news/who-pulls-the-plug-the-anthropic-fable-affair-and-what-it-means-for-everyone-else https://www.the-independent.com/tech/ai-sovereignty-nuclear-weapon-arms-race-trump-b3011359.html https://www.france24.com/en/live-news/20260630-cia-boss-compares-cutting-edge-ai-to-nuclear-weapons https://techgov.intelligence.org/research/an-international-agreement-to-prevent-the-premature-creation-of-artificial-superintelligence https://www.gov.uk/government/news/ai-firms-pioneering-drug-discovery-cheaper-supercomputing-and-more-get-first-backing-through-uks-sovereign-ai https://www.theregister.com/ai-and-ml/2026/07/07/mps-tell-brit-government-sort-out-your-tech-sovereignty-or-get-left-out-in-the-cold/5267076