Trump signs downsized AI order after weeks of reversals
Trump signed a downsized AI executive order requiring voluntary 30-day government review instead of the 90-day version he rejected last month.
Trump issued an executive order Tuesday addressing AI cybersecurity threats but with less stringent requirements than a previous draft. The final version asks some AI companies to submit powerful models for voluntary government review 30 days before public release, down from 90 days in the May draft that Trump abruptly rejected hours before a planned signing ceremony. The order also directs the Treasury Department to establish a cybersecurity clearinghouse with AI developers and critical infrastructure owners, creates a classified benchmarking process for assessing national security implications of advanced AI models, and tells the Pentagon to secure its networks within 30 days. The shift reflects tensions within Trump's administration between those seeking AI innovation to compete with China and those concerned about risks from models like Anthropic's Mythos, which researchers say can discover security vulnerabilities in widely used systems.
What HN community is saying
Comments debate whether the order is substantive or performative. Top responses question the mechanics of a 30-day review process and whether government agencies can meaningfully assess threats in that timeframe. Commenters express skepticism that the review is actually security-focused rather than ideologically motivated, citing Trump's July 2025 "stop woke AI" executive order that regulated LLM ideology. A lower-ranked concern notes this expands censorship frameworks rather than establishing genuine safeguards. One commenter provides context that the UK's AISI has conducted similar evaluations for years and links to NIST's comparable unit (CAISI). There is concern the order may primarily benefit incumbent AI companies by creating moats and coordination cover. Overall skepticism dominates about whether any of this meaningfully improves security.