Report from the New York Times
In Brief – The Trump Administration is weighing a shift in policy to allow federal agencies to review advanced AI models, with officials discussing an executive order to establish an AI working group composed of government leaders and tech executives that would explore a potential safety review system. The discussions mark shift in administration policies that rolled back Biden-era AI directives focused on “safety” requirements and emphasized rapid AI innovation and deregulation to compete with China. A key catalyst behind the shift was the development of Anthropic’s powerful model, Mythos, which reportedly has significantly improved capabilities to identify software vulnerabilities and engineer cyberattacks. The developer chose not to release the model publicly but instead shared it with major industry groups to allow them to identify and address risks. Administration officials are considering a range of ideas, including a regime that would give the government early access to advanced models for evaluation without authority to block their release, as many AI industry leaders continue to express concerns about regulatory burdens slowing innovation.
Context – The second Trump Administration’s AI policy is built on several firm beliefs, starting with AI leadership being a national imperative to compete with China, deregulation promoting the necessary level of investment and innovation, and a complete rejection of “Woke AI” in all its forms. When Vice President Vance hectored EU officials about AI policy that is “handwringing about safety”, or the White House Director of Science and Technology Policy said that the US “totally” rejects global governance of AI, he called out “ideological, risk-focused obsessions, such as climate or equity.” Not cybersecurity or bio and chemical weapons. AI regulation popping up in states to address “science fiction” fears about how AI might impact retail prices, discrimination, or mental health, often squarely hits the Woke guardrail, and polling showing Republicans and Democrats equally occupied by AI fears hasn’t changed the administration’s perspective yet.