A White House official confirmed on Monday that the Trump administration has finalized the voluntary AI evaluation framework. The initiative, announced as part of an executive order in June, aims to balance AI innovation with security concerns. While the White House did not disclose details of the framework, it plans to meet with industry partners to discuss its implementation.
What’s in the framework
The executive order directed the creation of a voluntary program for AI companies to submit powerful new models to the government for evaluation. This would allow the government to test frontier models, which are at the leading edge of AI development, for potential risks. One such model, Mythos from Anthropic, was already shared with partners for cybersecurity testing before wider deployment. AI developers will have the ability to voluntarily submit new models to the federal government up to 30 days ahead of their public release. The White House will then vet their cyber capabilities according to a classified benchmarking system and share the AI models with federal agencies and trusted corporate partners. Open models will reportedly be excluded.
The administration is working with companies to identify these high-risk models and access them for up to 30 days before public release. The voluntary nature of the program remains key. The order explicitly avoids imposing regulation that could hinder innovation.
The risks at stake
The government worries that these frontier models could uncover software vulnerabilities in critical systems. In the wrong hands, they might be used for cyberattacks or other malicious activities. The administration is testing whether a voluntary approach, rather than heavy-handed regulation, can mitigate these risks. Recent incidents, such as OpenAI and Anthropic discovering their AI models had unknowingly bypassed controls and hacked into third-party services during internal testing, have heightened officials' concerns. The House Committee on Homeland Security sent a letter to OpenAI CEO Sam Altman last week requesting that he brief lawmakers about how one of the company’s AI agents breached the platform Hugging Face.
Some argue that the secretive process will give an advantage to larger companies. “They're essentially creating an entrenchment program for the big AI model providers, which are now considered the most frontier,” says a person familiar with the White House’s discussions with AI labs, who requested anonymity to discuss confidential matters. “This creates an economic incentive program for critical infrastructure just to use them and leaves out smaller startups.”