Trump Signs AI Executive Order Seeking Early Access to Frontier Models

TLDR

• Trump signed an executive order seeking early federal access to advanced AI models.
• AI companies can voluntarily submit models for benchmarking before wider public release.
• The order focuses on models with advanced cyber capabilities and security risks.
• Federal agencies may review selected models up to 30 days before release.
• Anthropic’s Claude Mythos and Project Glasswing remain central to current AI security debates.

President Donald Trump signed an executive order on Tuesday targeting advanced artificial intelligence models before public release. The order asks AI companies to share certain models with the federal government through a voluntary review process. It focuses on models with advanced cyber capabilities and gives agencies a role in early testing.

Trump Order Sets Voluntary AI Review Process

The order asks AI developers to join a federal benchmarking process before wider model releases. It directs officials to assess advanced cyber capabilities and decide whether models meet covered frontier model standards.

The order seeks model access up to 30 days before companies release them more broadly. It also lets the government help select trusted partners that receive early access. The White House framed the process as voluntary and limited in scope. The order states that it does not create “mandatory governmental licensing” for new AI models.

Trump signed the order in private after delaying an earlier signing event with tech executives. He told reporters then that he disliked certain parts of the previous version. The final order carries fewer details than earlier proposals discussed in Washington. However, it still gives federal agencies a formal role in reviewing high-capability AI systems.

Agencies Get Cybersecurity Tasks Under New AI Directive

The order outlines several deadlines for agencies to prepare guidance and directives. It also directs the Department of Defense to prioritize cyber defense across its information systems. Federal officials will review models that may support advanced software vulnerability discovery. They will also examine whether those systems require controlled access before wider release.

The directive comes after Anthropic drew attention with the Claude Mythos Preview earlier this year. The model showed strong performance in finding software weaknesses and security flaws. Anthropic limited Mythos access through Project Glasswing, its cybersecurity program for selected partners. In our most recent report, the company has expanded Project Glasswing to about 150 additional organizations.

Those partners include groups in power, water, healthcare, communications, and hardware. Anthropic requires each partner to meet security standards before gaining model access. The company reported that Project Glasswing partners found more than 10,000 serious security flaws. It is also estimated that a major cyberattack could affect more than 100 million people.

Anthropic Dispute Adds Pressure Around Frontier AI

The order lands during a busy period for leading AI developers and public markets. In another report by Blockonomi, Anthropic confirmed on Monday that it had confidentially filed for an initial public offering. OpenAI has also prepared for a possible offering this year, according to the report. Meanwhile, SpaceX may reach the public market first through a planned debut.

The technology industry has played a central role in White House AI policy discussions. Elon Musk, Mark Zuckerberg, and David Sacks reportedly pushed against an earlier order draft.

Sacks previously served as the White House crypto and AI czar before the role ended. He remains a key ally of Musk and a known voice in AI policy debates. The order also follows meetings between Anthropic and senior Trump administration officials. Those meetings included White House Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent.

However, Anthropic still faces a dispute with the Department of Defense. The agency labeled the company a supply chain risk before the Mythos release. That designation limits defense contractors from using Anthropic technology in work with the agency. Anthropic sued the Trump administration to reverse the designation, and the case remains ongoing.

The post Trump Signs AI Executive Order Seeking Early Access to Frontier Models appeared first on Blockonomi.