Skip to content
Decaying

White House Weighs Mandatory Pre-Release AI Model Vetting — A Policy Reversal

Read in other languages: 한국어日本語
AI May 6, 2026 By Insights AI 1 min read 17 views Source

The Proposal

The Trump administration is in early discussions about an executive order requiring government review of powerful AI models before public release. Under options being considered, NSA, the Office of the National Cyber Director (ONCD), and the Director of National Intelligence would serve as oversight bodies. Importantly, review would not give the government power to block a model's release.

A Notable Policy Reversal

The administration revoked Biden's AI safety executive order in January 2025. The reported catalyst for this rethink is Anthropic's Mythos model, which demonstrated the ability to identify and exploit decades-old cybersecurity vulnerabilities. NEC Director Kevin Hassett confirmed May 6 that an executive order is under active study; a White House official called media characterizations "speculation."

Proposed Mechanics

  • Required model safety documentation
  • Structured red-team testing for cybersecurity exploits
  • Evaluation of dual-use risks
  • Mandatory disclosure of capabilities exceeding defined thresholds

NIST's CAISI separately announced expanded voluntary testing agreements with Google DeepMind, Microsoft, and xAI — a parallel track that doesn't require new legislation.

Source: Tom's Hardware

Share: Long

Related Articles