US blocks Anthropic's most powerful AI models worldwide - G7 negotiates 'trusted partners' access for allies
What it really says
On June 12, 2026, US Commerce Secretary Howard Lutnick issued an export control directive ordering Anthropic to block access to its latest AI models Fable 5 and Mythos 5 for all foreign nationals, regardless of whether they are inside or outside the United States. Even foreign employees of Anthropic itself are affected. Since there is no reliable technical method to verify nationality at the API level, Anthropic disabled both models globally for all users at 5:21 PM ET on June 12. The official justification cited a tip that Mythos 5 could be jailbroken to bypass its safety guardrails. Anthropic disputed this assessment, stating that the jailbreak in question was a narrow, single-instance exploit that would only unlock specific cybersecurity capabilities of Mythos, and that comparable capabilities exist in OpenAI's GPT-5.5, which is not subject to similar restrictions. Less powerful Claude models such as Claude Opus 4.8 remained unaffected. At the G7 summit in Evian-les-Bains on June 17, the situation took center stage. Sam Altman (OpenAI), Dario Amodei (Anthropic), Demis Hassabis (Google DeepMind), and Arthur Mensch (Mistral) attended a working lunch with heads of state and government. A 'trusted partners' framework was discussed that would grant vetted allies access to advanced US AI models. French President Macron said he expected progress on restoring access. German Chancellor Friedrich Merz emphasized that the potential of these technologies should be available to all countries, but that Europe also needed to catch up.
Our assessment
This incident warrants a yellow rating because it highlights legitimate structural concerns without being acutely threatening to everyday use. The good news: less powerful AI models like Claude Opus 4.8 or GPT-5.5 remain available, meaning the vast majority of practical AI applications continue to work normally. G7 nations are also actively working on a diplomatic solution. The legitimate concern: Washington has demonstrated that it can shut off global access to leading AI models overnight. The Centre for European Policy (CEP) calls the measure 'more of a geopolitical signal than a necessary security measure.' For European companies that have built critical business processes on US AI, this is a wake-up call. The fact that Anthropic itself questions the US government's justification and points out that comparable capabilities exist in unrestricted models suggests the decision was at least partly geopolitically motivated. For end users, this means: AI tools are not disappearing, but dependence on individual providers and countries carries risks that should be taken seriously.
Relevance for Germany
This development is significant for Germany for several reasons. First, numerous German companies - from startups to DAX corporations - use AI services from Anthropic, OpenAI, and Google. The ban shows that such access can be restricted at any time for political reasons. Second, Chancellor Merz personally negotiated for model access at the G7 summit while emphasizing that Europe needs to build its own capabilities, underscoring the federal government's growing awareness of digital sovereignty. Third, the Bundestag passed the AI Implementation Act just on June 11, implementing the EU AI Regulation nationally and establishing the Bundesnetzagentur as the supervisory authority. However, the Anthropic incident shows that regulatory control alone does not protect against the geopolitical risks of AI dependency. Fourth, the EU is advancing the construction of its own AI infrastructure with the Cloud and AI Development Act (CADA) from June 3, 2026 - an initiative that has gained new urgency from the export ban.
Fact check
Core information is confirmed by multiple independent sources. Anthropic confirmed the suspension in an official statement, documenting the exact time (June 12, 5:21 PM ET) and scope of the directive. Euronews, heise online, and US News report consistently on the G7 discussions on June 17 in Evian. Chancellor Merz's and President Macron's statements are documented by Reuters and US News. The CEP assessment as a 'geopolitical signal' comes from an analysis published on June 16, 2026. The fact that other Claude models such as Opus 4.8 remained unaffected is confirmed by both Anthropic and heise online.
Source
- • https://www.euronews.com/my-europe/2026/06/17/ai-takes-centre-stage-at-g7-as-western-fears-over-us-kill-switch-get-real
- • https://www.anthropic.com/news/fable-mythos-access
- • https://www.cep.eu/eu-topics/details/us-access-ban-on-anthropics-fablemythos-5-more-of-a-geopolitical-signal-than-a-necessary-security-measure.html
- • https://www.usnews.com/news/world/articles/2026-06-17/g7-leaders-vow-closer-ties-on-ai-as-they-hash-out-trusted-partners-scheme
- • https://www.heise.de/en/news/US-government-forces-shutdown-of-Anthropic-s-AI-Fable-5-and-Mythos-5-11331146.html