The proposal requires developers to prove their systems cannot facilitate complex cyberattacks before securing lucrative government contracts.
White House officials currently face rising pressure to address risks from advanced models like Anthropic’s Mythos.
The group recommends that the U.S. Center for AI Standards lead the development of these mandatory requirements.
Under this plan, Congress would establish a permanent enforcement office within the U.S. Department of Commerce.
The mandates would target major firms spending over $100 million annually on high-level model training.
Companies generating at least $500 million from AI-driven products would also face these strict oversight measures.
Existing voluntary agreements with OpenAI, Google, and xAI currently provide only limited oversight for federal agencies.
This federal initiative mirrors California’s reporting thresholds, which the state successfully enacted during the previous year.
