Deutsch
English
Español
Français
Nederlands
Português
The United States and the European Union are in active discussions to establish a 'trusted partners' framework for granting allied nations privileged access to cutting-edge AI models, according to sources familiar with the talks. The initiative comes in the wake of Washington's recent decision to restrict foreign access to Anthropic's most advanced AI systems, sparking concerns in Brussels about technological sovereignty and the weaponization of American AI capabilities.
The proposal was discussed by US Commerce Secretary Howard Lutnick with European diplomats on the sidelines of the G7 summit in Évian-les-Bains, France, which runs from June 15 to 17, 2026. The plan aims to create a tiered access system where close US allies receive preferential access to frontier AI models from American companies such as Anthropic, OpenAI, and Google, while maintaining national security safeguards against adversaries like China.
Background: The Anthropic Restrictions
On June 12, 2026, the Trump administration issued an export-control directive ordering Anthropic to block foreign nationals from accessing its two most advanced models — Claude Fable 5 and Claude Mythos 5 — citing national security concerns over a potential jailbreak technique that could exploit cybersecurity vulnerabilities. Anthropic complied but abruptly suspended access for all users globally on June 13, calling it a misunderstanding and working to restore access.
The move sent shockwaves through allied nations. Prior to the ban, approximately 150 organizations in more than 15 countries had access to Mythos under Anthropic's Project Glasswing, spanning healthcare, communications, power, and water sectors. The Anthropic AI model restrictions effectively cut off foreign researchers, hospitals, and critical infrastructure operators overnight.
'This is a wake-up call,' said French politician Bruno Retailleau. 'Nations dependent on foreign technology can be unplugged overnight.' British MP Al Carns noted that UK researchers, companies, and hospitals using the models were abruptly cut off, calling the situation 'unacceptable for a sovereign nation.'
What Is the Trusted Partners Framework?
The proposed 'trusted partners' arrangement would designate certain countries — likely G7 members and other close allies — as eligible for privileged access to advanced US AI models. The framework is being discussed as a mechanism to balance national security concerns with the strategic need to maintain allied technological capabilities, particularly in cybersecurity defense against state-sponsored threats.
Key Elements of the Proposed Framework
- Tiered Access: Countries or companies designated as 'trusted partners' would receive preferential access to frontier AI models that are otherwise restricted for foreign nationals.
- Cybersecurity Focus: The primary use case cited is strengthening allied cyber defenses against rivals like China and Russia, leveraging models like Anthropic's Mythos 5 — described as 'the strongest cybersecurity model globally.'
- Industry Participation: AI executives from Anthropic (CEO Dario Amodei), OpenAI (CEO Sam Altman), and Google are expected to participate in a working lunch on technology issues at the G7 summit.
- Voluntary Framework: The US executive order on AI (EO 14409, signed June 2, 2026) explicitly states it does not create mandatory government licensing or preclearance requirements for AI models, suggesting the trusted partners plan would be a cooperative arrangement.
European Reaction and Strategic Implications
The European Commission is examining the practical consequences of the Anthropic directive. A Commission spokesman said emergency measures should not discriminate against partners, signaling discomfort without formally challenging the US order. The episode has accelerated calls for European technological sovereignty and reduced reliance on US AI platforms.
French minister Benjamin Haddad characterized the decision as 'a turning point in the geopolitical struggle for AI.' Former French Prime Minister Édouard Philippe likened AI to critical infrastructure that nations must control domestically. The EU AI Act regulatory framework is being closely watched as Europe tightens its own AI governance while investing €200 billion in domestic computing capacity.
Dutch far-right leader Geert Wilders demanded the Netherlands accelerate its own AI development, while Jordan Bardella emphasized AI as a major issue of national sovereignty, urging support for French AI projects like Mistral AI.
Impact on Global AI Governance
The trusted partners discussions represent a pivotal moment in the geopolitics of artificial intelligence. The US approach under the Trump administration has been to promote AI innovation by slashing bureaucratic constraints while selectively restricting access to the most powerful models on national security grounds. This dual strategy — encouraging domestic innovation while controlling foreign access — creates a new paradigm for global AI technology competition.
Key statistics underscore the stakes: Anthropic, valued at approximately $965 billion as of May 2026, holds 34.4% of business AI adoption, ahead of OpenAI's 32.3%, according to Forbes. Its Fable 5 model scored 80.3% on the SWE-bench Pro software engineering benchmark, with Stripe reportedly compressing months of engineering work into days using the platform.
For enterprise customers worldwide, the situation changes risk calculations around AI procurement and continuity planning, particularly in banking, healthcare, and government sectors that had begun integrating Anthropic's models into critical operations.
FAQ
What is the US-EU trusted partners plan for AI?
The trusted partners plan is a proposed framework discussed at the G7 summit in June 2026 that would grant select allied nations privileged access to advanced US AI models from companies like Anthropic and OpenAI, balancing national security concerns with strategic cooperation.
Why did the US restrict access to Anthropic's AI models?
The Trump administration ordered Anthropic to block foreign nationals from accessing its Fable 5 and Mythos 5 models on June 12, 2026, citing national security concerns over a potential jailbreak technique that could exploit cybersecurity vulnerabilities.
Which AI models are affected by the restrictions?
Anthropic's Claude Fable 5 (the most capable widely released model with safety classifiers) and Claude Mythos 5 (an unrestricted version for vetted cybersecurity partners) were both suspended globally on June 13, 2026.
How has Europe reacted to the US AI restrictions?
European leaders have called the restrictions a 'wake-up call,' with France, the UK, the Netherlands, and others demanding accelerated domestic AI development and technological sovereignty to reduce dependence on US platforms.
What is the G7's role in AI governance?
The G7 summit in Évian-les-Bains is discussing AI regulation, infrastructure, and the trusted partners framework, with executives from Anthropic, OpenAI, Google, and Mistral AI participating in working sessions on technology policy.
Follow Discussion