NextFin News - The European Union is moving to "intensify" high-level discussions with the U.S. administration regarding the regulation and oversight of advanced artificial intelligence models, specifically those possessing sophisticated cyber capabilities. The diplomatic push follows the April release of Anthropic’s Mythos model, a system whose advanced autonomous hacking and defensive abilities have triggered a security debate spanning from Brussels to Washington. According to a Commission spokesperson speaking to CNBC, the EU is now seeking broader technical cooperation to manage the risks of AI-powered cyberattacks while navigating a complex web of U.S. export and security restrictions.
The friction centers on Project Glasswing, Anthropic’s initiative to roll out Mythos to a select group of organizations. While the U.K.’s AI Security Institute has been granted preview access, the EU’s AI Office remains excluded. Sources familiar with the matter told CNBC that Anthropic informed the Commission that any expansion of access to non-U.S. government entities requires explicit permission from the White House. This bottleneck highlights the growing tension between the U.S. government’s desire to maintain a "maximum calculus" of security and the EU’s ambition to enforce its own landmark AI Act on models operating within its borders.
U.S. Treasury Secretary Scott Bessent, representing the administration of U.S. President Trump, characterized the current strategy as a delicate balancing act between innovation and national security. Bessent, who has historically advocated for a "pro-growth, America-first" economic policy, noted on Thursday that the administration is working "very closely" with AI labs to ensure the U.S. maintains its lead over China. His stance reflects a broader administration priority: treating advanced AI as a strategic asset that must be shielded from foreign oversight, even that of traditional allies, to prevent the leakage of dual-use capabilities that could be weaponized by adversaries.
This protective stance is not without its critics. Some industry analysts suggest that by restricting the EU’s access to Mythos, the U.S. risks creating a regulatory vacuum or a "splinternet" of AI standards. However, the White House’s opposition to Anthropic’s plan to expand access to 70 additional organizations—as reported by the Wall Street Journal—suggests that security concerns currently outweigh the benefits of international regulatory alignment. For the EU, the inability to audit Mythos presents a direct challenge to the implementation of its AI Act, which mandates transparency for "systemic risk" models.
The "cyber moment" warned of by Anthropic CEO Dario Amodei appears to have arrived, forcing a confrontation between the borderless nature of AI development and the rigid boundaries of national security policy. As the EU seeks to mutually recognize standards with the U.S., the outcome of these intensified talks will likely determine whether the next generation of AI is governed by a unified Western framework or a fragmented system of bilateral permissions. For now, the White House remains the ultimate gatekeeper of the world’s most potent cyber AI, prioritizing the preservation of a technological edge over the demands of global regulatory harmonisation.
Explore more exclusive insights at nextfin.ai.
