The European Union is escalating its engagement with the U.S. administration regarding the most advanced artificial intelligence models, particularly those possessing “cyber capabilities.” This intensified dialogue comes amid growing concerns over the potential misuse of powerful AI for malicious cyber activities.
A significant catalyst for these discussions was the recent announcement of Anthropic’s formidable “Mythos” model. The unveiling of this cutting-edge AI in April triggered a palpable sense of urgency among governments and the corporate sector, sparking widespread apprehension about AI-driven cyberattacks. The Wall Street Journal reported in the same month that the White House had expressed reservations about Anthropic’s initial plans to broaden access to this highly capable model.
Anthropic, however, indicated on Thursday that “Mythos-class” models are slated for customer release in the coming weeks. This timeline underscores the accelerating pace of AI development and deployment.
The U.S. administration maintains that it is actively collaborating with leading AI laboratories to achieve a delicate equilibrium between fostering innovation and ensuring robust safety measures. This strategic approach is crucial as the U.S. strives to maintain its technological preeminence in the global AI race, especially in competition with China.
Initially, Anthropic provided preview access to the Mythos model through an initiative known as Project Glasswing, extending it to a select cohort of companies and organizations. However, the AI research firm has yet to grant preview access to the EU, its AI office, or any government entities outside the United States, with the exception of the U.K.’s AI Security Institute.
The Wall Street Journal had previously reported that Anthropic had proposed expanding access to Mythos to approximately 70 additional companies and organizations. This proposal reportedly faced opposition from U.S. administration officials due to significant security considerations.
“Cybersecurity is a shared priority, and we have agreed to mutually recognize our respective standards in this area,” a Commission spokesperson, Thomas Regnier, told CNBC, referencing ongoing discussions between the Commission and the U.S. administration. “In parallel, we are now expanding technical discussions with Anthropic and other model developers who have notified their latest models to the AI office.”
Sources familiar with the confidential deliberations between the EU and Anthropic revealed that during discussions concerning preview access to Mythos, Anthropic informed the Commission that the bloc would first need to secure permission from the U.S. administration. These individuals requested anonymity as they are not authorized to speak publicly on the matter.
These same sources added that Anthropic had communicated to the Commission that the White House’s objection was not specifically to sharing Mythos with the EU, but rather a broader stance against the company sharing such advanced models with non-U.S. governments.
When approached for comment, the White House press office directed CNBC to recent briefing remarks made by Treasury Secretary Scott Bessent. Bessent emphasized that the U.S. administration is engaged in “very close” collaboration with AI labs.
“They’ve been excellent partners, and we are going to get a solution that solves for maximum calculus,” Bessent stated, reinforcing the U.S.’s position as the world’s AI leader, with China in second place. “We want to make sure that we keep that lead. So, we are working on the exact calculus between innovation and safety, and we want to optimize for that.”
Anthropic itself has acknowledged that models with Mythos’ advanced capabilities necessitate stringent cyber safeguards prior to their general release. The company stated, “We’re making swift progress on developing these safeguards and expect to be able to bring Mythos-class models to all our customers in the coming weeks.”
Earlier this month, Anthropic CEO Dario Amodei sounded a significant warning, highlighting a critical six-to-12-month window to address tens of thousands of software vulnerabilities identified by the Mythos model. He expressed concern that failure to act swiftly could allow Chinese AI advancements to surpass current U.S. capabilities.
Anthropic declined to provide further comment.
Original article, Author: Tobias. If you wish to reprint this article, please indicate the source:https://aicnbc.com/22243.html
