Trump is set to sign an AI oversight order this week as MAGA security concerns intensify regarding frontier-model laboratories.

      The draft executive order, anticipated to be released as soon as Thursday, establishes a voluntary framework for a 90-day pre-release model disclosure to the federal government, engaging critical infrastructure providers such as banks early in the process. Both Steve Bannon and Amy Kremer have advocated for a more stringent, mandatory approach.

      President Donald Trump is likely to sign an executive order on AI oversight shortly, according to a report from Reuters on Wednesday, in response to increasing demands from parts of his political base for a stricter security review of advanced AI systems.

      The proposed order creates a voluntary system where AI developers would grant the federal government access to powerful models 90 days prior to their public launch, as well as earlier access for operators of critical infrastructure, including banks.

      The political impetus moving the order toward final approval is notably unique in its alignment. U.S. News & World Report explains that a coalition within the broader MAGA base, featuring former Trump adviser Steve Bannon and right-wing political organizer Amy Kremer, has been urging the administration to enforce mandatory security reviews for the most advanced AI models.

      Bannon argues that the releases of Anthropic’s Mythos and OpenAI’s GPT-5.5-Cyber have changed the cyber-threat landscape significantly, which the federal government must address.

      The voluntary framework represents a careful compromise within the West Wing. Advisors with industry ties, such as David Sacks, have voiced concerns about mandatory disclosures on competitiveness grounds, while national-security advocates aligned with Bannon and Kremer have taken the opposite stance.

      The 90-day voluntary structure is a product of this internal discussion. According to the draft language available, the order does not stipulate civil or criminal penalties for those who choose not to participate, although opting out may be designated as a public stance that the government could highlight.

      The model disclosure framework is designed in response to a specific cyber-threat context. Anthropic’s prior promise to share Mythos findings with partner governments, along with the recent briefings regarding Mythos's cyber capabilities, has set a precedent that the executive order formalizes within the U.S.

      The Trump administration frames the voluntary system as granting federal cybersecurity agencies (CISA, NSA’s Cybersecurity Collaboration Center, and parts of the FBI) the same pre-public release period that companies have already provided to foreign partner intelligence services concerning the Mythos timeline.

      The provision for pre-release access to critical infrastructure is expected to be implemented rapidly. According to Vorys’s legal analysis of the broader White House AI governance strategy, the critical infrastructure category includes banks, energy providers, telecommunications companies, and large healthcare networks.

      The 90-day window is designed to enable these operators to assess their defenses against new frontier models before they are widely deployed to the public.

      The order is being introduced during a politically charged period. The recent Trump-Xi Beijing summit regarding AI regulations and Nvidia H200 export licensing has established a bilateral framework that influences the U.S.-China AI policy landscape.

      The executive order set to be signed this week serves as a domestic complement, tailored to the security review identified as a necessary institutional component in the bilateral discussions.

      Local media coverage has focused on the pressure from the MAGA base that has shaped the timing of the order, with the release aligned with political messaging about AI safety as much as with any specific cyber threat situation.

      The White House has not released the final text of the order prior to the signing. Details such as the list of voluntary participants, the federal coordinating body for the 90-day disclosure framework, the identified critical infrastructure recipients for pre-release access, and the timeline for the first model release under the framework remain unconfirmed.

      The administration has indicated that OpenAI, Anthropic, Google DeepMind, Meta, and Microsoft are all expected to take part from the outset. The next tangible milestone will be the first model disclosed under the voluntary framework, which the White House anticipates may occur before the end of Q3 2026.