Last Friday, the Trump administration ordered Anthropic to block foreign citizens from accessing its latest artificial intelligence models. The move has been viewed by European officials and policymakers as a concrete warning that Washington could abruptly cut off critical AI services, heightening concerns among U.S. allies about risks tied to technological dependency. The export ban targets Anthropic's Mythos and Fable models—software increasingly adopted by banks, law firms, and government agencies—and extends the logic of U.S. export controls on AI chips directly to model services themselves. During the G7 summit held in France, this restriction quickly became a focal point of discussion among allies. According to three diplomatic sources, G7 leaders discussed on Tuesday a proposal to allow certain “trusted partners” access to advanced AI models from American companies like Anthropic, potentially opening a pathway to circumvent U.S. restrictions on non-residents using cutting-edge AI systems. Sources said that during the opening dinner of the summit, representatives from multiple countries primarily discussed expanding access with U.S. Commerce Secretary Linda McMahon, noting that “trusted partners” could be either nations or corporations. White House officials stated that the Trump team remains in communication with allies and is committed to addressing national security concerns raised by Anthropic’s models.
Cybersecurity experts believe that Anthropic’s Mythos model, designed to identify code vulnerabilities, could significantly enhance the capacity to attack banking technology systems. The EU is seeking access to this model to study its implications. Prior to the Trump administration’s order, Anthropic had already opened access to Mythos for select organizations in over 15 countries, using it to scan computer systems across healthcare, communications, power, and water sectors for vulnerabilities. Executives from Anthropic, OpenAI, and Google are expected to attend a working lunch during the summit to discuss regulatory frameworks, AI infrastructure, and cybersecurity issues. This incident has also accelerated Europe’s push for indigenous alternatives. On Tuesday, French Prime Minister Gabriel Attal announced that France’s public sector will deploy tools based on domestic startup Mistral AI, while France’s domestic intelligence agency, DGSI, has already chosen local company Chapsvision to replace U.S. software firm Palantir, emphasizing that France must possess its own tools and cannot rely on the goodwill of partners who might arbitrarily cut off access to Anthropic’s models. Palantir stated that its multi-year contract with DGSI, renewed through 2025, remains valid. Meanwhile, G7 leaders, during their meeting in Évian, made AI a key agenda item, discussing both the opportunities and risks posed by AI—especially those relevant to the financial sector.
Image source: internet
Original article: toutiao.com/article/1868212026607691/
Disclaimer: The views expressed in this article are solely those of the author(s).