The draft executive order, expected as early as Thursday, sets up a voluntary 90-day pre-release model-disclosure framework with the federal government, with critical-infrastructure providers including banks brought in early. Steve Bannon and Amy Kremer have been pressing for a harder, mandatory line.
President Donald Trump is expected to sign an executive order on AI oversight as soon as Thursday, Reuters reported on Wednesday, under mounting pressure from parts of his political base who want tighter security review of frontier AI systems.
The draft order, sets up a voluntary framework under which AI developers would provide pre-release access to powerful models to the federal government 90 days before public launch, alongside pre-release access for critical-infrastructure operators including banks.
The political pressure that has moved the order to the signing table is unusual in its direction. U.S. News & World Report’s read of the order describe a coalition inside the broader MAGA constituency, including former Trump adviser Steve Bannon and right-wing political organiser Amy Kremer, who have been pressing the White House to require mandatory security review of the most capable frontier models.
Bannon’s framing, on the coverage, is that the launches of Anthropic’s Mythos and OpenAI’s GPT-5.5-Cyber have shifted the cyber-threat surface in a direction the federal government cannot afford to ignore.
The 💜 of EU tech
The latest rumblings from the EU tech scene, a story from our wise ol' founder Boris, and some questionable AI art. It's free, every week, in your inbox. Sign up now!
The order’s voluntary structure is a deliberate middle-ground compromise inside the West Wing. Industry-facing advisers including David Sacks have pushed back against the mandatory-disclosure framing on competitiveness grounds; Bannon-Kremer-aligned national-security voices have pushed in the opposite direction.
The voluntary-90-day-window structure is the result of that internal negotiation. The order does not, on the available draft language, impose civil or criminal penalties for non-participation, though declining to participate is framed as a public posture the government can call out.
The model-disclosure framework is calibrated against a specific cyber-threat surface. Anthropic’s earlier commitment to share Mythos findings with partner governments, including the prior week’s FSB-side briefings around Mythos cyber capabilities, established the pattern the executive order codifies on the US-domestic side.
The Trump administration’s framing, on the coverage, is that the voluntary regime will give federal cybersecurity agencies (CISA, NSA’s Cybersecurity Collaboration Center, and elements of the FBI) the same pre-public-release window the company has already given to foreign-partner intelligence services on the Mythos cycle.
Critical-infrastructure pre-release access is the part of the order most likely to be operationalised quickly. Vorys’s legal-analysis read of the broader White House AI-governance plan has the critical-infrastructure category covering banks, energy operators, telecoms providers and large healthcare networks.
The 90-day window is intended to allow those operators to stress-test their own defensive posture against the new frontier models before they reach broader public deployment.
The order arrives inside a politically charged window. The Trump-Xi Beijing summit on AI guardrails and Nvidia H200 export licensing established the bilateral track that frames the US-China AI policy environment.
The executive order under signature this week is the domestic-side complement, calibrated to the security review that the bilateral track has flagged as the missing US institutional layer.
Local-affiliate coverage has emphasised the MAGA-base pressure that has produced the timing, with the order’s release calibrated to the political-base messaging cycle around AI safety as much as to any specific cyber-threat event.
The White House did not publish the final order text ahead of the signing. The voluntary-participation list, the federal-agency coordinating body for the 90-day disclosure framework, the named critical-infrastructure recipients of the pre-release access, and the timeline for the first model disclosed under the framework have not been confirmed.
The administration has, signalled that OpenAI, Anthropic, Google DeepMind, Meta and Microsoft are all expected to participate from launch. The next visible proof point will be the first model disclosed under the voluntary framework, which the White House timeline suggests could land before the end of Q3 2026.
Get the TNW newsletter
Get the most important tech news in your inbox each week.