Covered frontier models get a classified threshold and a 30-day pre-release window
The June executive order on advanced AI innovation and security directs a classified benchmarking process for cyber capability, with the NSA Director determining when a model crosses into covered frontier model status. Developers may volunteer models for government access 30 days before release. The order also expressly forbids creating any licensing or pre-clearance requirement.
Both halves of that are deliberate. The order builds a designation mechanism with real teeth on the assessment side, then closes the door on the enforcement side — nothing in it authorises mandatory licensing, pre-clearance or permitting for developing, publishing or releasing a model.
The result is an unusual instrument: a capability threshold that carries no automatic legal consequence, paired with a voluntary framework under which developers hand over 30 days of pre-release access subject to confidentiality, insider-risk and IP protections. Treasury is separately directed to stand up an AI cybersecurity clearinghouse for vulnerability identification and remediation.
Whether it works turns entirely on the benchmarking. A classified threshold is only meaningful if the harness measuring against it is sound — and a government-authored cyber benchmark just leaked a model onto the open internet. The order's reach is also narrower than its early drafts; earlier reporting tracked a postponement and a scaled-back scope before signature.
The White House — Promoting Advanced Artificial Intelligence Innovation and Security → · Federal Register — Promoting Advanced Artificial Intelligence Innovation and Security → · Wiley — New AI executive order addresses frontier models and cybersecurity vulnerabilities → · Lawfare — White House releases executive order on AI →