Sources: WH is preparing to order US agencies to partner with AI companies on cybersecurity; the EO wouldn't require pre-release model testing by the government

The Trump administration is preparing to order US agencies to partner with artificial intelligence companies to protect networks …

Bloomberg 2026-05-09

Context & Ripple Effects

Related coverage shows the White House weighing several AI-security measures at once: rules for national-security agencies, an oversight working group, and potential restrictions on companies that impede government model use.

This reported cybersecurity order would place agency–AI-company collaboration alongside a separately discussed voluntary early-access framework, while stopping short of mandatory government testing before release.

First-order effects

US agencies and AI companies would be pushed toward formal cybersecurity partnerships, making model providers more directly involved in protecting government networks.
The reported choice not to mandate pre-release testing preserves a lighter-touch approach for developers than the vetting option discussed in related coverage.

Second-order effects

AI vendors seeking government work may need to build clearer pathways for agency access and cybersecurity collaboration, even if model testing remains voluntary rather than compulsory.
Agencies would have to manage vendor concentration as they adopt these tools, consistent with the related policy memo's reported emphasis on avoiding reliance on a single provider.

Third-order effects

If the measures advance together, US AI policy could develop through procurement, national-security access, and voluntary cooperation rather than a centralized pre-release approval regime.
That model may increase the government's operational dependence on a small set of capable AI suppliers, making multi-vendor resilience and the terms of government access enduring policy questions.

The trend: The administration appears to be shaping AI governance around national-security deployment and voluntary industry cooperation rather than centralized model regulation.

Discussion

@andrewcurran_ Andrew Curran on x
Looks like government approval for frontier model releases may be off the table. [image]
@neil_chilson Neil Chilson on x
This sounds very reasonable: [screenshot: “...A draft executive order from President Donald Trump would revamp existing cybersecurity information-sharing programs to include Al companies and address threats posed by the emerging technology, said the people, who spoke on condition…
@adamthierer Adam Thierer on x
glad to hear we're getting back to sanity on federal AI policy and discarding extreme ideas like an “FDA for AI.” These are more reasonable steps 👉
@eastland_maggie Maggie Eastland on x
Scoop: The White House is preparing an executive order on AI security that omits mandatory model testing and emphasizes voluntary participation by AI companies for cyber defense efforts. W/ @cmsub for @business https://www.bloomberg.com/...
@mikeshepard4 Mike Shepard on x
NEW: The White House is preparing to order US agencies to partner with artificial intelligence companies on cybersecurity, though it would stop short of requiring government approval for cutting-edge AI models —> @eastland_maggie & @cmsub for @business https://www.bloomberg.com/.…
@dareasmunhoz Diego Areas Munhoz on x
This appears to be the security EO that's been floated for some time now. It incorporates elements of the AI Action Plan. And it's not the vetting EO Hassett and others have been talking about that has scared lots in the industry https://www.whitehouse.gov/... [image]

Chronicles