EU seeks to 'intensify' talks with U.S. on advanced AI models amid Mythos concerns

The European Union is eager to enhance its discussions with the U.S. regarding cutting-edge AI models, particularly those with potential cyber capabilities. This move comes in response to rising concerns stemming from Anthropic’s recently launched Mythos model, which has stirred considerable anxiety among governments and businesses alike over the risks of AI-driven cyberattacks. Reports from the Wall Street Journal indicate that the White House has expressed opposition to Anthropic’s plans to broaden access to Mythos. The administration is currently collaborating closely with AI developers to find a balance between fostering innovation and ensuring safety, as the U.S. aims to maintain its competitive edge over China in the global AI landscape. Initially, Anthropic provided access to Mythos to a limited selection of companies and organizations under an initiative named Project Glasswing. However, entities in the EU and other non-U.S. governments, except for the U.K.'s AI Security Institute, have yet to receive any preview access to evaluate the model. According to the WSJ, Anthropic had suggested allowing an additional 70 companies and organizations to utilize Mythos, yet U.S. officials have voiced concerns regarding security implications. Thomas Regnier, a spokesperson for the Commission, stated, "Cybersecurity is a shared priority, and we have agreed to mutually recognize our respective standards in this area," highlighting ongoing talks with U.S. officials. During the discussions about granting preview access to Mythos, Anthropic conveyed that the EU would need to seek U.S. approval first. A source familiar with the negotiations, speaking on condition of anonymity, revealed that while the White House is not opposed to sharing Mythos with the EU specifically, it generally resists sharing with governments outside the U.S. In remarks made to CNBC, Treasury Secretary Scott Bessent emphasized the administration’s close collaboration with AI labs, stating, "They've been excellent partners, and we are going to get a solution that solves for maximum calculus." He reiterated the U.S.'s position as the foremost leader in AI technology, with China trailing behind, and underscored the importance of maintaining that lead while optimizing the balance between innovation and safety. Anthropic's CEO, Dario Amodei, warned that there is a critical six- to twelve-month window to address numerous software vulnerabilities identified by the Mythos model before Chinese AI advancements potentially catch up. Anthropic has opted not to comment further on these developments.