The Trump administration’s decision to block foreign access to Anthropic’s most advanced AI models, Fable 5 and Mythos 5, has rapidly escalated into a transatlantic and Indo-Pacific diplomatic flashpoint. Anthropic, valued in the tens of billions of dollars and expected to list on the stock exchange in 2026 alongside SpaceX and OpenAI, found itself at the centre of a national security storm after the White House identified a jailbreak vulnerability that could allow users to bypass safety guardrails. Officials gave the company just 90 minutes to suspend all non-US usage, according to the Financial Times, and have since issued an ultimatum: stop disputing the severity of the flaw and start patching it. Viewed from Washington, the subsequent negotiations over a formal framework for assessing AI security risks represent a belated effort to build government infrastructure that can keep pace with technology, though the administration and Anthropic CEO Dario Amodei remain at odds over the real-world danger posed by the jailbreak.

The export controls have reverberated sharply across the G7. At a summit on 17 June, French President Emmanuel Macron and Indian Prime Minister Narendra Modi voiced fears that the United States could unilaterally sever access to critical AI systems, leaving allies and strategic partners vulnerable. Their concerns prompted a discussion of a “trusted partner” mechanism that would guarantee continued access to advanced models for a coalition of like-minded nations, reducing dependence on Washington’s policy shifts. European and Asian officials increasingly view American dominance in frontier AI not as a stable commercial offering but as a geopolitical lever, exposing the fragility of global AI supply chains.

The private sector has moved swiftly to insulate itself. JPMorgan Chase and Goldman Sachs have instructed employees outside the United States—notably in Hong Kong—to cease using Anthropic’s Claude models, citing licensing terms that now restrict non-US access. The preemptive bans, reported by the Financial Times, reflect a broader compliance anxiety: even without explicit government orders, the legal ambiguity surrounding export-controlled AI is prompting multinationals to wall off their own workforces. This fragmentation risks creating a two-tier AI environment, where cutting-edge capabilities are reserved for American users and a handful of approved jurisdictions.

Domestically, the controversy has fuelled a parallel debate over who should own and benefit from artificial intelligence. Senator Bernie Sanders introduced the American AI Sovereign Wealth Fund Act, which would impose a one-time tax payable in stock, effectively transferring 50 per cent equity of firms with over $200 million in annual AI revenue to a public fund. The fund, projected to be worth $7 trillion, would distribute annual dividends exceeding $1,000 to every American. While the bill stands no chance of passage in a divided Congress, it has sharpened the political conversation around inequality and democratic control, with Sanders warning against allowing “a handful of Big Tech oligarchs” to dictate the future.

Amid the turmoil, Anthropic is pursuing a dual track: pledging closer cooperation with the Commerce Department and promising faster resolution of security issues, while simultaneously forging ahead with product development. The company this week unveiled a major update to Claude Design, enabling centralised brand governance and automated design compliance—a signal that it intends to maintain commercial momentum. Analysts in London note that the standoff, and the broader scramble it has triggered, marks a turning point where AI governance is no longer a theoretical exercise but a live contest between national security, corporate autonomy, and international equity. The outcome of the Anthropic framework talks will likely set a precedent for how the world’s most powerful AI models are shared—or withheld.