Hajra Naz
President Donald Trump signed a new executive order on Tuesday. It asks major artificial intelligence companies to give the federal government early access to their most advanced AI models. Officials say the aim is to spot cybersecurity risks and reinforce protections for critical infrastructure before wider release.
The decision comes as concern grows around powerful AI systems, including Anthropic’s Mythos. The model has drawn attention from government agencies and financial institutions. Reports suggest it can detect and exploit cybersecurity weaknesses at speeds that worry many experts.
Under the order, AI developers are encouraged to share certain advanced models with the government up to 30 days before those systems are provided to other partners. The review process remains voluntary. The administration believes early access could help agencies prepare for emerging cyber threats.
The signing had been expected nearly two weeks earlier. It was pulled hours before the planned White House event.
A draft reviewed in May by CNN proposed a 90-day review window. That version sparked debate inside the AI industry. Several companies pushed for a shorter period, arguing that development cycles move too fast for long review timelines.
Read More: Melania Trump Introduces Futuristic Humanoid Robot at White House
A person familiar with internal discussions said senior Trump advisers and cabinet officials met on Monday. Those talks helped finalize the revised version signed this week.
A White House spokesperson said the order reflects cooperation between government and industry. The goal, the spokesperson said, is to support innovation while addressing national security concerns.
The order also directs national security agencies to set up a cybersecurity clearinghouse. Officials say it would improve coordination and speed up information sharing on cyber threats.
It also states that participation does not create a licensing system. There is no requirement for government approval before releasing AI models.
Several major AI companies were involved in discussions with the White House. Anthropic was among them. The company has previously faced scrutiny from the Pentagon over its handling of safety controls in sensitive systems.
Read More: OpenAI is reportedly developing a smartphone where AI agents replace apps entirely
OpenAI also supported the move. Chief Global Affairs Officer Chris Lehane called it a step that aligns safety with innovation. He said both need to move together to maintain U.S. leadership in AI.
Lehane added that AI safety rules should continue to develop through public institutions. He pointed to input from experts, policymakers, and industry groups as part of that process.
CNN reported that OpenAI CEO Sam Altman is expected to meet White House officials and members of Congress on Wednesday. Details of the agenda were not shared.
Microsoft President Brad Smith also welcomed the order. He said it supports innovation while addressing security risks facing the public.
The Trump administration had previously taken a lighter regulatory approach to AI. That position began to shift after concerns around Anthropic’s Mythos model gained attention. The company said the system could uncover cybersecurity flaws at unusual speed.
Experts have warned that advanced AI could increase both the scale and speed of cyberattacks. Supporters of the order argue that early government review may help reduce those risks.
Anthropic and OpenAI have already given select governments and organizations access to advanced models for security testing. Anthropic also said Mythos will roll out to all customers in the coming weeks.
The Commerce Department’s National Institute of Standards and Technology had earlier announced plans for companies to share unreleased AI models for security evaluation. That notice has since been removed from its website.
Read More: OpenAI and Microsoft have torn up their exclusivity agreement
The executive order had been set for signing on May 20. Executives from major tech firms were expected at the White House event. Some were already on their way when it was postponed.
Trump later said he delayed the order because he disagreed with parts of the draft. He said some provisions risked slowing AI development. He added that he did not want the policy to become a barrier for progress in the sector.
