Trump to sign AI oversight this week as MAGA security pressure mounts on model labs at border
The draft executive order, expected to take effect Thursday, establishes a 90-day model disclosure framework with the federal government, and critical infrastructure providers including banks that have been brought forward. Steve Bannon and Amy Kremer have been pushing a hard, mandatory line.
President Donald Trump is expected to sign an AI oversight order as soon as Thursday, Reuters reported on Wednesday, under growing pressure from parts of his political base calling for a tougher security review of border AI systems.
The draft order, sets out a voluntary framework in which AI developers will provide pre-release access to powerful models to the federal government 90 days before public launch, as well as pre-release access to operators of critical infrastructure including banks.
The political pressure that has brought the order to the signing table is unusual in its direction. US News & World Report’s reading of the order describes a coalition within the MAGA region, including former Trump adviser Steve Bannon and right-wing political activist Amy Kremer, who have been pushing the White House to demand a mandatory review of border security models.
Bannon’s argument, on the cover, is that the introduction of Anthropic’s Mythos and OpenAI’s GPT-5.5-Cyber has moved the cyber threat into an area where the federal government cannot ignore it.
💜 for EU tech
The latest talk from the EU tech scene, a story from our genius founder Boris, and some incredible AI art. It’s free, every week, in your inbox. Register now!
The voluntary structure of the order is a deliberate compromise within the West Wing. Industry experts including David Sacks have rejected the mandatory disclosure framework on competition grounds; Bannon-Kremer-aligned security voices have gone the other way.
The 90-day window structure is the result of that internal negotiation. The order, in the draft language available, does not include civil or criminal penalties for non-participation, although refusal to participate is listed as a public stance the government can declare.
The model’s exposure framework is tailored to a specific cyber threat environment. Anthropic’s previous commitment to share Mythos findings with partner governments, including last week’s FSB-side reports on Mythos’ cyber capabilities, established a pattern for the executive order to include on the US-domestic side.
The framework of the Trump administration, in this case, is that the voluntary regime will give the government’s cybersecurity agencies (CISA, the Cybersecurity Collaboration Center of the NSA, and elements of the FBI) the same window of previous releases that the company has given to foreign intelligence services in the Mythos cycle.
Pre-release access to critical infrastructure is part of the order that is likely to be used quickly. Vorys’ legal analysis read through the White House’s broader AI governance program has a category of critical infrastructure that includes banks, electronics companies, telecom providers and major healthcare networks.
The 90-day window is intended to allow those operators to assess their defense posture against new frontier models before they reach wider public deployment.
The order comes within a politically charged window. The Trump-Xi Beijing conference on AI guardrails and the granting of an export license for Nvidia H200 have established a bilateral track that sets the environment for US-China AI policy.
The executive order signed this week is to complement the domestic side, which has been assessed in a security review that the bilateral track has been marked as a layer of the missing US facility.
The coverage of the local company emphasized MAGA-base pressure that produced the time, with limited order release in the politically motivated message cycle surrounding AI security as much as any specific cyber threat event.
The White House did not publish the final text of the order before it was signed. The voluntary participation list, the federal agency’s coordinating body for the 90-day disclosure framework, the critical infrastructure recipients named for access to the previous release, and the timeline for the first model disclosed under the framework have not been confirmed.
Executives have indicated that OpenAI, Anthropic, Google DeepMind, Meta and Microsoft are all expected to participate from launch. The next point of tangible evidence will be the first model revealed under the voluntary framework, the timeline of the White House suggests that it could arrive before the end of Q3 2026.
