White House unveils pared-back AI executive order
The White House has released a revised executive order addressing artificial intelligence, significantly shortening the mandatory review period for AI model releases from 90 days to 30 days, responding to industry pressure. The order focuses on collaboration between government and industry to bolster cybersecurity, particularly concerning critical infrastructure and tracking cyber threats. It establishes a cybersecurity clearinghouse to facilitate vulnerability scanning and prioritization of patching efforts.
The revised executive order represents a shift in approach from the initial, more stringent proposal. The key change – reducing the review period – was driven by concerns from AI industry leaders, including Google, OpenAI, and Anthropic, who argued the longer timeframe hindered innovation. The order now emphasizes a faster, collaborative process for assessing AI models, particularly ‘frontier’ models, which have recently raised cybersecurity concerns due to their potential for autonomous vulnerability discovery. This shift reflects a broader strategy of deemphasizing regulation and prioritizing a more agile response to the rapidly evolving AI landscape.
The order also establishes a framework for government-industry collaboration on cybersecurity, including the creation of an AI cybersecurity clearinghouse led by the Treasury Department. This clearinghouse will facilitate the sharing of vulnerability information and prioritize patching efforts. Furthermore, the Office of the National Cyber Director, CISA, and OMB will work together to identify and allocate grant funding for advanced AI vulnerability detection. The signing of the order occurred behind closed doors, following a similar cancellation of a public event for the previous iteration, highlighting the administration's sensitivity to industry concerns.