President Donald Trump signed an executive order Tuesday directing federal agencies to ask artificial intelligence companies for early access to their models, the culmination of several weeks of internal deliberations inside the White House.
After scuttling an earlier version of the executive order at the last minute, Trump approved an order that aims to have artificial intelligence companies voluntarily share information about “frontier models” with government agencies.
The deliberations among Trump officials centered on balancing federal agencies’ need to get ahead of the potentially disruptive impacts of AI with the administration’s desire to take a light touch in regulating tech companies. A similar version of the plan was pulled in May after David Sacks, the president’s former AI czar, raised concerns about its potential impact on the industry, but the administration appears to have overcome those concerns.
The biggest difference is that agencies now have a 30-day period to preview the new AI models rather than a 90-day period, as originally planned.
Trending
“The United States continues to lead the world in Artificial Intelligence (AI) because of the enormous talent and innovation of our AI industry, and because we refuse to stifle this innovation with overly burdensome regulation,” Trump’s order states. “Advanced AI capabilities make our Nation stronger, but also introduce new national security considerations that require coordinated action across executive departments and agencies (agencies), and components.”
The order also included the line: “Nothing in this section shall be construed to authorize the creation of a mandatory governmental licensing, preclearance, or permitting requirement.”
The White House order gives the Cybersecurity and Infrastructure Security Agency and the National Institute of Standards and Technology 30 days to create a standard to decide which models will be covered by the order.
The executive order also directs the Defense Department to deploy cybersecurity measures to protect against vulnerabilities created by advanced AI models. It similarly directs CISA to coordinate with AI companies to address cybersecurity vulnerabilities in “critical infrastructure,” such as utilities and rural hospitals.
Though the order is voluntary, AI safety advocates have celebrated the White House’s recognition that the technology poses potential safety risks. An ongoing high-profile clash over the deployment of AI between Anthropic and the Pentagon illustrated the divergent interests between the government and private firms over the technology.
“This is an immense shift over the tone the White House has struck on AI over the past … 12 months, where AI safety was not just an afterthought, but it was actually rejected as something that we need to be concerned about at all,” Chris MacKenzie, a spokesperson for Americans for Responsible Innovation, an AI safety advocacy group, told NOTUS.
President Joe Biden signed an executive order in 2023 that included mandatory information-sharing agreements with the federal government. Tech groups threatened to challenge the order, and Trump scrapped it on his first day back in office.
The Trump administration shifted its tone toward AI safety a few weeks ago, after Anthropic released its latest AI model, Claude Mythos, which alarmed the intelligence community and cybersecurity industry because of its advanced ability to find cybersecurity vulnerabilities.
The White House has included AI labs like OpenAI, Anthropic and Google in these voluntary efforts, according to multiple outlets. The White House has also briefed industry groups, some of which previously threatened to challenge Biden’s AI executive orders.
Authors
