US Executive Order Establishes Voluntary Pre-Release Review of Frontier AI Models (infosecurity-magazine.com)

• Voluntary pre-release cybersecurity review of frontier AI models by US government
• NSA, CISA, NIST to develop classified benchmark for 'covered frontier model' threshold
• No mandatory licensing or preclearance required
• Response to threat from models like Anthropic's Claude Mythos that exploit vulnerabilities at scale
• 30-day hardening of federal systems; creation of AI cybersecurity clearinghouse under Treasury

"President Trump signed an executive order on June 2 creating a voluntary framework for pre-release cybersecurity review of frontier AI models by the US government. Developers may submit models for up to 30 days of review before public release. The order explicitly rules out mandatory licensing. NSA, CISA, and NIST are tasked with building a classified benchmark to determine which models are 'covered.' The order also mandates a 30-day hardening of federal systems, directs CISA to issue binding directives, and creates an AI cybersecurity clearinghouse led by the Treasury Department. Industry reactions were supportive but skeptical of voluntary effectiveness."