EU Rejects US 'Security Risk' Claim on Anthropic, Warns Against AI Discrimination

Just days after their public launch, Anthropic was forced to abruptly disable its most advanced AI models, Fable 5 and Mythos 5, for all users globally. A US Commerce Department directive, issued on June 12, 2026, mandated that the company block access to any foreign national, including its own non-US employees, citing unspecified national security concerns . The European Commission has now firmly pushed back, rejecting the core justification for the order and warning against discrimination against allies .

"Europe Is Not a Security Risk"

The Commission's rejection is rooted in the view that the US action treats trusted partners as adversaries. The directive’s sweeping scope—barring access for all non-US citizens, whether inside or outside the United States—is seen in Brussels as disproportionate and fundamentally unwarranted for a long-standing security partner like the EU .

A blunt statement from EU Tech Sovereignty Chief Henna Virkkunen captured the official stance: "Europe does not pose a 'security risk' to the United States. Europe represents an economic opportunity, not a security risk. We are and will remain a trusted partner" . The Commission is formally "assessing the implications" of the export control and has explicitly warned that such measures "should not be discriminatory" .

The practical fallout was immediate. The directive did not just affect external customers; it also blocked Anthropic's own foreign-national employees in the US from working on the models, effectively forcing a global shutdown to ensure compliance .

A Wake-Up Call: EU Officials Decry Discrimination and an "AI Kill Switch"

The incident has been met with alarm across European capitals, framed as a "wake-up call" that exposes Europe's vulnerability to US-controlled AI infrastructure.

• Discrimination Warning: The European Commission’s primary complaint is that the Americans-only order punishes allies. A spokesperson stressed that measures impacting partners "should not be discriminatory" .
• Sovereignty in "Overdrive": The Register reported that the clampdown has sent EU sovereignty efforts "into overdrive," as officials view it as proof that Washington can unilaterally flip an "AI kill switch" on critical technology relied upon by European businesses and governments .
• French Alarm: French officials and lawmakers described the move as a hostile act, with Le Monde reporting that the incident has "worried" France and Europe, triggering fears of a US cut-off from essential AI tools—prompting fresh calls to support homegrown champion Mistral as a sovereign alternative .

The Legislative Response: Fast-Tracking the Cloud and AI Development Act

The political urgency created by the shutdown has breathed new life into pre-existing plans for European tech sovereignty. The Cloud and AI Development Act (CADA), a cornerstone of the Commission's AI Continent Action Plan announced just days before the Anthropic incident, is now being fast-tracked as a direct countermeasure .

• What CADA Does: The legislation aims to strengthen Europe's sovereignty in cloud and AI by building an integrated, EU-wide cloud ecosystem, reducing dependence on non-EU (chiefly US) infrastructure, and establishing a single cloud policy for public-sector procurement .
• Sovereignty by Procurement: A critical component of CADA is a four-tier sovereignty assessment that would, in practice, bar US hyperscalers from the most sensitive government contracts, citing jurisdictional exposure under US laws like the CLOUD Act as a formal disqualification .
• A New Urgency: The Anthropic shutdown is being used as Exhibit A for why Europe cannot rely on AI platforms that can be disabled at Washington’s discretion. The push now includes accelerated development of EU-native AI models, new procurement rules favoring sovereign cloud services, and discussions of potential retaliatory measures if future US controls block allied access to frontier AI .

The Bigger Picture: From Cooperation to Tech Independence

While the Commission has stated it will seek to "intensify" technical dialogues with the US on advanced AI models—particularly those with cyber capabilities—the trust deficit is clear . The incident has crystalized a consensus that technical autonomy is no longer a distant aspiration but an urgent necessity.

The European Technological Sovereignty Package, presented on June 3, 2026, which bundles CADA with a Chips Act 2.0, now has the political wind at its back . The message from Brussels is unmistakable: Europe will remain a trusted partner, but it will no longer be a captive user of technology it cannot control.