President Donald Trump signed an executive order on Tuesday asking artificial intelligence companies to provide their most advanced models to the federal government for security testing before they release them to the public. The order seeks voluntary participation from AI developers in a benchmarking process designed to assess whether models pose national security risks, marking a new approach to sovereign AI oversight in the United States.

What Does Trump's AI Order Actually Require?

The executive order asks AI companies, on a voluntary basis, to participate in a benchmarking process that evaluates a model's "advanced cyber capabilities" and determines whether it qualifies as a "covered frontier model." If a model meets that threshold, the order requests access to it up to 30 days before the company plans to release it more broadly to the public.

The order also enables the government to help select which "trusted partners" will receive early access to these models. However, the order explicitly states that "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," meaning companies cannot be forced to comply.

Trump signed the order privately, weeks after he postponed a signing ceremony with prominent tech CEOs because he "didn't like certain aspects of it," according to comments he made to reporters at the time. The Tuesday order is notably thin on specific details about how the benchmarking process will work or which agencies will oversee it.

Why Is the Tech Industry Paying Attention to This Timing?

The order arrives at a critical moment for the AI industry. On Monday, just one day before Trump signed the order, Claude developer Anthropic announced it had confidentially filed with the Securities and Exchange Commission for an initial public offering (IPO). Rival OpenAI is also preparing for a potential public offering later this year.

The timing is particularly significant given recent tensions between Anthropic and the Trump administration. The Department of Defense had labeled Anthropic a "supply chain risk" shortly before the company released Claude Mythos Preview, a model designed to identify security flaws in software. That designation prohibits defense contractors from using Anthropic's technology in their work with the agency. Anthropic has since sued the Trump administration to try to reverse that designation, and the litigation is ongoing.

Anthropic did expand its cybersecurity initiative called Project Glasswing on Tuesday, the same day Trump signed the order. The company had previously captivated government officials and Wall Street by announcing Claude Mythos Preview, which prompted high-profile meetings between Anthropic executives and senior Trump administration members, including White House Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent.

How Is the Tech Industry Responding to Government Oversight?

The tech industry has played a central role in shaping the White House's positions on AI policy. Venture capitalist David Sacks, a longtime ally of Elon Musk, served as the first crypto and AI czar before that role ended earlier this year. Last month, Sacks, along with Musk and Meta CEO Mark Zuckerberg, reportedly called the Trump administration to lobby against a prior version of the AI executive order that the president had prepared to sign.

The successful lobbying effort suggests that tech leaders have significant influence over how the administration shapes AI governance. The voluntary nature of Trump's final order appears to reflect industry concerns about mandatory government oversight, which many companies view as potentially burdensome to innovation and competitiveness.

Steps for Understanding Government AI Benchmarking Requirements

• Frontier Model Classification: Companies must first determine whether their AI models qualify as "covered frontier models" based on their advanced cyber capabilities, which will be assessed through the government's benchmarking process.
• Voluntary Participation Timeline: If a company chooses to participate, it must provide early access to models up to 30 days before planned public release, allowing the government time to conduct security assessments.
• Trusted Partner Selection: The government will help identify which external organizations can receive early access to models, creating a controlled rollout process before broader public availability.
• No Mandatory Licensing: Companies retain the right to decline participation, as the order explicitly prohibits the creation of mandatory governmental licensing or preclearance requirements for AI model development and release.

The order also directs the Department of Defense to prioritize the cyber defense of its information systems, suggesting that national security concerns about AI capabilities are driving the benchmarking initiative. The DOD's recent designation of Anthropic as a supply chain risk indicates that government agencies are already evaluating which AI companies and models pose potential security concerns.

Meanwhile, Elon Musk's SpaceX, which owns his AI lab SpaceXAI, is poised to beat both Anthropic and OpenAI to the public market, with a debut set to take place as soon as the following week that could value the company at well over one trillion dollars. This competitive landscape suggests that how the government approaches AI oversight could significantly influence which companies succeed in the coming years.