White House Artificial Intelligence Oversight Framework
Issued by
Executive Office of the President of the United States (White House)
- September 30, 2026 · Correction — Added what is confirmed: the framework is voluntary, covers only closed frontier models, and has not been published. (Cody Maxwell)
- September 30, 2026 · Correction — Corrected the date (presented 4 August 2026, not an effective date), removed binding review duties and a procurement-based scope, and noted it covers only frontier closed models (fact-check finding). (Cody Maxwell)
- October 1, 2026 · Correction — Rewrote the practical steps, newsletter hook, search description, and audience fields to match the entry's corrected content. (Cody Maxwell)
The White House finalized a voluntary framework for testing the most advanced AI models' cyber capabilities before release, under the June 2026 executive order. It presented the framework to AI companies on 4 August 2026 and has not published the text.
Applies To
Overview
The White House finalized this framework by 1 August 2026 and presented it to AI companies on 4 August 2026. It sets a voluntary process for reviewing advanced AI models before release. It implements the June 2026 executive order on frontier AI and cybersecurity. Participation is voluntary, and it covers only the most capable closed models, not open-weight ones. The administration has not published the text. Companies choose whether to take part. Enforcement mechanisms and specific penalty structures have not been publicly confirmed. The framework may interact with or build upon prior executive orders addressing AI safety and national security, including those issued in 2023 and 2025. It matters most to developers of the most capable closed models.
Key Requirements
- •Participation is voluntary; developers choose whether to submit models before release.
- •It covers only the most capable closed models, which the June 2026 order calls covered frontier models.
- •Reviews test models' cyber capabilities before release to partners and the public.
- •The text is not public; only participating companies see the details.
- •It sets no binding submission or review duties and no penalties.
What Your Organization Must Do
- →Confirm whether your firm develops the most capable closed models, the only ones in scope.
- →Treat participation as a business choice, since the framework sets no binding duties or penalties.
- →Ask White House contacts for the framework text, which has not been published.
- →Brief engineering teams on cyber capability testing before release, if your firm opts in.
- →Leave federal procurement contracts unchanged, because this framework imposes no deadline or vendor obligation.
- →Track the June 2026 executive order on frontier AI and cybersecurity for any later binding rules.
Playbook Guidance
Step-by-step implementation guidance for compliance teams.
Governance Controls
Operational controls that implement requirements from this regulation.
Frequently Asked Questions
- Does the White House AI Oversight Framework apply to private companies not contracting with the federal government?
- It is not tied to federal contracting. It covers developers of the most capable closed models, who can choose to submit them for review before release.
- What are the pre-deployment review requirements under the WH-AIOF and when do they take effect?
- Companies that opt in submit their most advanced models for cyber capability testing before release. The White House presented the framework on 4 August 2026; it has no effective date and its text is not public.
- What penalties does the White House AI Oversight Framework impose for non-compliance?
- None, because the framework is voluntary and sets no penalties for companies that do not take part.
- How does the WH-AIOF relate to the 2023 and 2025 executive orders on AI safety?
- It implements the June 2026 executive order on frontier AI and cybersecurity (EO 14409). EO 14110 of 2023 was revoked in January 2025, so the framework does not build on it.
- Should federal AI vendors update procurement contracts now or wait for the full framework text to be released?
- There is no deadline to meet. Federal AI vendors need not change contracts because of this framework, since it is voluntary and aimed at frontier model developers.
- Which types of AI models trigger review obligations under the White House AI Oversight Framework?
- Only the most capable closed models, which the June 2026 order calls covered frontier models. Open-weight models are excluded, and taking part is voluntary.