Donald Trump signed an executive order to create a voluntary framework for the federal government to vet powerful new AI models before they are released. Tuesday’s highly anticipated order represents an attempt by the president to tighten his grip on cybersecurity and national security threats posed by AI, tacking against his earlier deregulatory stance. But the voluntary nature of the framework shows that, while Trump has toed a more cautious line on AI than when he first took office last year, he is still reluctant to impose regulations on the tech industry.
Under the new guidelines, tech companies would be asked to share their AI models with the government for a voluntary review, up to 30 days before a public release. The Trump administration says doing so will allow them to improve national security, particularly with regards to cybersecurity.
The executive order stops short of imposing mandatory review requirements on tech companies building AI models, a rumored feature of earlier versions of the executive order. Some of Trump’s more hardline Maga supporters had been pushing him for a stricter process, while tech industry supporters advocated for keeping the reins loose.
Trump quietly signed Tuesday’s long-awaited executive order after postponing the original plan – worried that its initial text would have been too restrictive on tech companies. An earlier version of the order reportedly requested the models 90 days before their release. The postponement in late May was in part the result of Silicon Valley leaders’ influence on the White House, according to reports from multiple news outlets. Tech billionaires including Elon Musk, Mark Zuckerberg and former White House “AI czar” David Sacks had personally pushed Trump to reverse course in private phone calls.
“I didn’t like certain aspects of it, I postponed it,” Trump said in the Oval Office on 21 May. “We’re leading China, we’re leading everybody, and I don’t want to do anything that’s gonna get in the way of that lead.”
The final executive order states that the US “continues to lead the world in AI” and that it “refuse[s] to stifle this innovation with overly burdensome regulation”. That declaration – and the lack of a mandate – suggests Trump is prioritizing innovation over regulation, and returning closer to his initial stronger pro-growth stance. One of his first actions as president was to revoke a Biden-era executive order that established standards for safely developing AI.
“Nothing in this section shall be construed to authorize the creation of a mandatory governmental licensing, preclearance, or permitting requirement for the development, publication, release, or distribution of new AI models, including frontier models,” Tuesday’s executive order reads.
Under the new guidelines, developers would be able to engage the federal government to determine how risky their model may be, and whether it would benefit from the framework. If officials deem the model would, tech companies would provide the federal government with access to these models for a period of up to 30 days before public release.
The new guardrails come amid rising fears that the latest AI models can be dangerous, especially in the wrong hands. Anthropic’s Mythos, a model with advanced cybersecurity capabilities, has raised concerns among AI safety experts, governments and tech companies, for its ability to exploit vulnerabilities at an unprecedented scale in widely used software.
Last month, the Trump administration struck a deal with Microsoft, Google DeepMind and xAI to review early models of their new AI models before they are released. (The federal government recently removed details of that agreement from its website, although it’s unclear why.) The Center for AI Standards and Innovation (CAISI), part of the US Department of Commerce, already has similar deals with OpenAI and Anthropic. The federal government says this kind of information sharing is standard practice and important for national security, although some free speech advocates have warned that too much government control could lead to censorship.
The National Security Agency and the Department of Defense will help determine which AI models need government scrutiny, and the treasury department will play a key role in finding vulnerabilities in AI models. The Trump administration also directed the government to hire more cybersecurity and AI professionals, and ensure there are stronger cybersecurity systems at institutions running key infrastructure, like rural hospitals, community banks and local utilities. The executive order also mentions speeding up and prioritizing “the cyber defense of civilian federal government information systems” and establishing or expanding “federal programs and cybersecurity services that enhance AI-enabled defensive tools”.
Trump announced another AI-focused executive order in December aimed at preventing states from regulating AI, which created a federal taskforce to challenge states’ AI laws.
