The US government issued an executive order on June 2, 2026, directing federal agencies to develop a voluntary framework for AI developers to provide access to frontier AI models before their public release. This access period, lasting up to 30 days, aims to protect cybersecurity, intellectual property, and national security interests, according to medianama.com.
The order tasks the Secretary of War, Homeland Security, Treasury, and the Director of the National Security Agency with creating a classified benchmarking process to assess AI models' cyber capabilities. It defines criteria for designating 'covered frontier models' and encourages collaboration among agencies and AI developers to promote secure innovation and protect critical infrastructure. The directive emphasizes that companies are not required to obtain licenses or approvals before releasing AI models.
This move reflects growing US government concern over the rapid development of advanced AI technologies and their potential risks. By seeking early access to frontier AI models, officials aim to strengthen cybersecurity defenses and safeguard intellectual property. The order aligns with broader efforts to regulate AI while balancing innovation, similar to initiatives in other countries addressing AI governance and risk mitigation.
Anthropic, a leading AI company, publicly welcomed the executive order, signaling industry support for the government's approach. The classified benchmarking process is expected to be developed within 60 days, setting the stage for ongoing federal oversight of AI model releases.