AI Chiefs Call on US to Police Frontier Models

The titans of cutting-edge AI suddenly seem to be reading from the same script on one point: regulate us, and do it soon. In separate essays published over the last five weeks, OpenAI's Sam Altman, Google DeepMind's Demis Hassabis, and Anthropic's Dario Amodei each laid out strikingly similar blueprints for policing so-called "frontier" models—tools they say could pose national security risks if left unchecked, per Axios. All three back independent pre-release testing, a centralized system to set rules and certify compliance, and a leading role for Washington in shaping oversight.

Where they diverge is on who holds the big red button. Amodei favors a federal agency able to halt model launches outright. Hassabis pushes for a self-regulatory group, funded by the industry but overseen by the government. Altman argues for a US-led international body that could condition market access on safety standards. The convergence comes as the Trump administration, publicly pro-deregulation, has already quietly stepped in to curb Anthropic's Fable and Mythos models and OpenAI's GPT-5.6 over cyber concerns, per Axios. The US is now "actively engaging with global partners to advance responsible, pro-innovation AI development," a State Department rep tells CNBC.