AI security order.
Trump signs voluntary AI safety order, balancing innovation with cybersecurity and national defense.
President Donald Trump signed a scaled-back executive order creating a voluntary framework under which developers can share advanced AI models with the federal government up to 30 days before public release. The order is intended to address cybersecurity risks highlighted by Anthropic’s “Mythos” and directs agencies to bolster cyber-defense capabilities, including establishing an “AI cybersecurity clearinghouse” with industry and critical infrastructure operators to coordinate scanning and prioritize remediation and patching. Within 60 days, the Treasury Department, the NSA, CISA, and NIST (with White House officials) must build a classified benchmarking process to assess advanced cyber capabilities and determine which models qualify as “covered frontier models.” The order states the review is not mandatory and bars any interpretation that would enable government licensing, preclearance, or permitting requirements for releasing new AI models. The order follows an earlier, more stringent draft that Trump delayed and then canceled amid industry and adviser concerns that longer review periods could slow innovation, while also aiming to preserve U.S. competitiveness in the AI race with China.
Key Points
The executive order sets out an explicit policy rationale: promoting “AI innovation and security” by working with the private sector to “modernize government and private sector information systems,” “harden them against external threats,” protect “intellectual property from exploitation and theft,” and “cultivate” U.S. “AI-enabled capabilities.”
It was prompted by Anthropic’s “Mythos” controversy: Anthropic said in April it was limiting Claude Mythos’s release because it was “too adept at finding software and cybersecurity vulnerabilities,” and the order follows ongoing federal conversations with the company; the startup has also “confidentially filed for an initial public offering” and said it would soon release Mythos-level models to customers.
The shift to the final 30-day window was described as a compromise between competing proposals: a leaked/original draft called for up to 90 days of pre-release access, while tech companies pushed to cut that figure to just 14 days.
David Sacks—portrayed as a key adviser behind efforts to prevent mandatory licensing—argued publicly that “unnecessary regulation is the biggest threat to innovation in America,” and said winning the AI race requires clearing “bureaucratic hurdles” from policymakers.
Sentiment
Mostly NegativePublishers
91
Articles
559
Reach
650
