US ban on Anthropic models: Cybersecurity experts demand lifting of export restrictions

The critics' reasoning

The authors explicitly classify the USA's decision as a political instrument. They speak of one of the most far-reaching interventions in AI policy ever. The block is not justified by comprehensible security-related benefits but is based on the assumption of a unique selling point that does not exist in practice.

The background to the controversy is the prior history involving the Fable 5 and Mythos 5 models. Anthropic had released Fable 5 last Tuesday as a content-restricted variant of the new Mythos 5. If a so-called classifier classified a query as sensitive, it was not Fable that answered the question, but the older Claude Opus 4.8 model. The aim was to avoid risky outputs from the more powerful model.

Background of the block

On Friday, Anthropic then unexpectedly blocked access to Fable 5 and Mythos 5 – for all users worldwide. The reason given was that there had been a possibility of circumventing Fable 5's protective mechanisms. According to a notice from the company, this could also have happened with "safe, normal content." This could explain why not only the restricted Fable 5 but also the full Mythos 5 had to be blocked.

The USA ordered the shutdown of Anthropic's AI models Fable 5 and Mythos 5, among other reasons, because of the suspicion that a group with ties to the Chinese government had access to Mythos 5. The block therefore affects not only distribution within the USA but also export to other countries, including Europe.

AI expert Anselm Küsters from the Freiburg Centre for European Policy sharply criticized the approach. "Jailbreaks als Rechtfertigung für eine weltweite Abschaffung zu nutzen, ist problematisch", he judged. He pointed out that security vulnerabilities in AI systems are not a new phenomenon and should be addressed with targeted measures rather than with a blanket ban.

Reactions from Germany and Europe

German business is also raising the alarm. Ralf Wintergerst stated: "Deutschland und Europa sind beim Zugang zu den stärksten KI-Modellen vom Wohlwollen der USA abhängig". This dependence is strategically risky, as European companies and research institutions depend on access to state-of-the-art AI systems in order to remain competitive.

Alexander Dobrindt warned of the geopolitical consequences of the current development. Dobrindt warns: KI-Rückstand macht Deutschland verwundbar. He sees the Federal Republic in a phase in which technological sovereignty and security-policy stability are increasingly interlinked. Without its own high-performance AI infrastructure, Germany would remain in a weak position vis-à-vis the USA and China.

In the political debate, the ban is increasingly being interpreted as a risk to the competitiveness of the West as a whole. Critics argue that a blanket export restriction on American AI models plays into the hands of adversarial actors. If European and other international users no longer have access to US models, they could increasingly turn to Chinese alternatives. This would achieve the opposite of what security policy intends.

Risk to Western competitiveness

The signatories of the open letter see a contradiction between declared security logic and actual effect. They call on the USA to withdraw the export restrictions and instead rely on differentiated security standards that enable responsible handling of the models without preventing their global availability.

According to the initiators, the signatories include representatives of established security firms as well as independent researchers. They emphasize that their criticism is not directed against regulation of AI systems as such, but against an approach based on blanket bans rather than on comprehensible technical criteria.

It is noteworthy that cybersecurity professionals from the Asia-Pacific region are also supporting the letter. They see restricted access to Western AI models as a weakening of the global cybersecurity community, which depends on the exchange of information about vulnerabilities and defense strategies.

Outlook and next steps

The USA has so far not publicly responded to the letter. Observers expect the debate to gather pace in the coming weeks, particularly as the European elections and the upcoming G7 consultations on AI governance approach. The EU Commission is also monitoring the development with a view to its own regulation of the AI market.

Should the ban remain in place, this is likely to further fuel the discussion about European AI sovereignty. Initiatives for own data centers and European AI models would then gain political momentum. At the same time, experts warn that without access to the strongest existing systems, Europe could lose touch with technological progress.

A statement from Anthropic's management is expected in the coming days. The company has so far been reticent and referred to ongoing talks with US authorities. Observers interpret the open letter as a signal that the technical community regards the conflict not merely as an internal US matter, but as a global question of AI governance.