The October 2023 executive order on AI was the most comprehensive US federal AI policy action to date. Tracking what actually happened through implementation is a useful exercise for enterprise compliance teams — the headline coverage often overstates enforcement and understates the durable institutional changes.
What moved: the NIST AI Safety Institute got funded and staffed. The AI RMF generative AI profile (covered separately) is a direct output. Federal agencies produced AI use inventories — the first time there’s been a systematic accounting of how federal agencies are using AI. Several agencies published procurement guidance that, for the first time, requires AI vendors to provide detailed documentation about model provenance, training data, and evaluation results.
What stalled: the reporting requirements for frontier model developers around advanced AI systems. The threshold triggers (computation thresholds in training runs) were set conservatively in the original order, and follow-on rulemaking to update them has been slow. The voluntary commitments from major labs remain voluntary.
What changed with the administration transition: the Biden AI executive order was revoked in the early days of the Trump administration and replaced with a narrower order prioritizing “AI dominance” framing over the safety-first framing of the original. The practical effects: the NIST AI Safety Institute continues its technical work; the voluntary commitments from labs became less prominent in administration communications; and the regulatory appetite for mandating safety evaluations at frontier labs decreased.
What’s still moving: state-level AI regulation (California, Texas, Colorado with various trajectories), the EU AI Act implementation timeline, and international coordination through the G7 and OECD AI Policy Observatory.