A look at a copy of the AI EO that President Trump was expected to sign on May 21; the unsigned EO emphasized that government AI reviews would be voluntary
The draft also includes language aimed at bad actors. It directs the attorney general to enforce the Computer Fraud and Abuse Act and …
Context & Ripple Effects
The unsigned draft sits within an administration arc that began by revoking Biden-era AI policies described as barriers to innovation and advancing an AI Action Plan aimed at rolling back prior measures on bias and cybersecurity.
Related coverage shows the May draft was followed by a signed, scaled-back order focused on AI cybersecurity threats and subjecting AI to less scrutiny than the abandoned version. The draft’s voluntary-review approach and its direction to enforce the Computer Fraud and Abuse Act illustrate that narrower posture.
First-order effects
- AI developers and federal agencies face less immediate compliance pressure from the draft’s contemplated review process because the reviews were voluntary and the order was not signed as drafted.
- The administration’s near-term enforcement emphasis shifts toward misuse by “bad actors,” using existing Computer Fraud and Abuse Act authority rather than creating a centralized AI-review regime.
Second-order effects
- Companies building or deploying AI gain a policy signal that federal intervention may center on cybersecurity and unlawful conduct rather than broad predeployment review, affecting how they prioritize government-facing risk programs.
- The scaled-back signed order leaves a more fragmented policy field: federal restraint on AI oversight can intensify disagreements over whether states should fill perceived governance gaps, an issue already dividing Republicans in related coverage.
Third-order effects
- If this approach persists, US AI governance is likely to develop through targeted use of existing cybercrime and security authorities rather than an AI-specific regulator—though the eventual scope will depend on later executive and legislative action.
- The episode reinforces a competitive policy model in which reducing perceived regulatory barriers is treated as part of AI industrial strategy, while safety obligations are concentrated around clearly defined security threats.
The trend: This is one data point in the Trump administration’s shift from broad AI safeguards toward lighter-touch, security-focused enforcement intended to preserve room for AI development.