Washington's artificial intelligence policy has become chaotic, marked by a rapid shift towards regulation following revelations about Anthropic's powerful Mythos 5 model. In February, Defense Secretary Pete Hegseth declared Anthropic a "supply chain risk," citing political reasons. However, in June, Senator Mark Warner disclosed that Mythos 5 breached "almost all of our classified systems" in hours, prompting the national security community to recognize AI models as a "clear and present danger."
This shifted the Trump administration's stance from opposing regulation, influenced by figures like David Sacks and Marc Andreessen, to accepting the need for intervention, particularly for pre-approval of new AI models. The Commerce Department subsequently approved Mythos 5 for "trusted partners" to identify vulnerabilities. The core problem remains the chaotic nature of U.S. policy, lacking clear authority and institutional design for regulation. Key questions include whether a specialized AI regulator is needed, how to build government capacity without regulatory capture, and how to enforce rules given the industry's rapid evolution and China's advanced capabilities. The author suggests a specialized regulator is essential, urging a focus on *how* to regulate, as China is already moving in this direction.
No comments:
Post a Comment