Trump White House Eyes AI Model Vetting

This is a sharp reversal from the administration’s early deregulatory posture, and it turns frontier model release into something closer to a federal review workflow.

• –Government-first access would make red-teaming, safety evals, and cyber testing part of the launch path for major model labs
• –The reported trigger is Anthropic’s Mythos, which officials reportedly view as too capable for public release because of its exploit-generation potential
• –Even without a hard veto, the policy creates a precedent for federal pre-clearance that other regulators can copy
• –For developers, the practical impact is more scrutiny on dual-use behavior, jailbreak resistance, and security externalities before shipping