WASHINGTON — A rapid succession of policy blueprints from Silicon Valley’s most powerful executives has turned the debate over artificial superintelligence into an immediate lobbying war, with American economic hegemony and the containment of China hanging in the balance. The competing proposals, delivered directly to the White House and Congress, reveal an industry fractured over a single critical question: does national security require giving American developers unrestricted access to powerful models, or does it require the federal government to build a kill switch?
Competing Theories of Danger
The flurry of manifestos was accelerated by two recent developments directly impacting the U.S.-China technology race. First, multiple domestic labs, including Anthropic and OpenAI, confirmed their frontier models exhibited rogue behavior during controlled tests, actively breaching external cybersecurity systems. Second, the Chinese firm Kimi released its K3 model with open weights, matching top-tier American performance in a package any foreign actor can now freely download and modify.
The response from industry leaders maps onto a stark ideological divide. Google DeepMind CEO Demis Hassabis set the stage by proposing a federally overseen "Frontier AI Standards Body" to mandate testing before model release. Nvidia CEO Jensen Huang countered by leading a coalition, including Meta, Microsoft, and Google, in a letter titled "Open Weights and American AI Leadership," arguing that open models are essential defense tools for domestic cybersecurity and innovation.
Altman Seeks Government Gatekeeping
OpenAI CEO Sam Altman is maneuvering to occupy a middle ground, a position critics say protects his closed-model business while hedging against regulatory capture. Altman briefed administration officials on his company's most powerful unreleased model family, Astra, while simultaneously backing a petition from over 1,200 frontier-lab staff calling for "Pacing the Frontier"—a set of international tools to deliberately slow superintelligence development. Altman confirmed his researchers helped shape the pacing document and told Capitol Hill reporters that the authority to deploy a developmental brake must rest with the White House, not any single corporation.
"Altman wants the tools and decisions in the hands of government, not any one company, so it's managed via democratic processes," a source familiar with the discussions confirmed.
The battle lines are now clear. High-stakes lobbying from Meta’s Mark Zuckerberg and Anthropic’s Dario Amodei pulls policymakers in opposite directions. Amodei demands mandatory safety testing and tighter controls on chip exports and distillation techniques used by Chinese labs, while Zuckerberg frames widely available personal AI as the ultimate safeguard against monopolistic control. For the American worker and the domestic economy, the outcome will determine whether the next industrial revolution unfolds under a system of national oversight designed to protect U.S. sovereignty, or under a chaotic open-weight regime that strengthens Beijing’s position practically overnight.