EU cybersecurity agency meets Anthropic over Mythos AI access

EU Cybersecurity Agency Enisa Set to Meet Anthropic Over Mythos AI Risks

The European Union Agency for Cybersecurity, known as Enisa, is set to meet with U.S. artificial intelligence company Anthropic in San Francisco on Thursday to discuss the cybersecurity implications of the firm’s advanced Mythos model, according to reports confirmed by the EU. The meeting, arranged at Anthropic’s invitation, comes as G7 leaders wrap up discussions on a “trusted partners” scheme that could grant allied nations access to powerful American AI systems for cyber defense.streetinsider

G7 Backs ‘Trusted Partners’ Framework

The Enisa meeting coincides with the conclusion of the G7 summit in Evian-les-Bains, France, where leaders from June 15 to 17 discussed a proposal to provide select “trusted partners” with access to advanced AI models from U.S. firms like Anthropic. French President Emmanuel Macron said on June 17 that he expected progress in the coming weeks on broadening access to leading U.S. AI models.reuters

The “trusted partners” could include both countries and corporations, Reuters reported, and would allow them to develop stronger cybersecurity defenses against rivals like China. In a joint statement, G7 leaders tasked finance officials, regulators, and cybersecurity experts with assessing how frontier AI models could affect financial stability, productivity, and labor markets.straitstimes

Why Mythos Matters

Anthropic first revealed its Mythos-class models in April, triggering emergency responses from central banks and intelligence agencies globally. The company claims Mythos possesses an ability to identify and exploit vulnerabilities in the software systems governing banks, power grids, and government operations. Access was initially restricted to 11 U.S.-based partner organizations, with only the United Kingdom granted access outside the United States.nytimes

Since then, the EU has been in active discussions with Anthropic about obtaining access. A European Commission spokesman confirmed in June that the bloc had held “several productive meetings” with Anthropic, and that Enisa was the body expected to gain access once an agreement was reached.whtc

A Broader Expansion Underway

On June 9, Anthropic publicly released Fable 5, the first consumer-facing model from the Mythos class, while simultaneously expanding the unrestricted Mythos 5 model to roughly 200 organizations across more than 15 nations through its Project Glasswing initiative. Anthropic has described Mythos as possessing “the strongest cybersecurity capabilities of any model in the world”.theguardian

Thursday’s meeting in San Francisco could mark a step toward Enisa joining that group, potentially giving Europe’s primary cybersecurity authority direct access to one of the most consequential AI systems yet developed.