What happened
President Trump cancelled an event to sign an executive order on AI safety this week. The order would have given the government the power to test frontier AI models before their public release. Ars Technica reported the cancellation came hours before the scheduled signing on Thursday, after several top AI CEOs declined to attend. Trump told reporters he felt the order could have been an 'innovation blocker' and did not want to slow the US lead over China.
How the room's reading it
The move splits the room. The tech industry broadly lobbied against the order, fearing safety testing could delay model launches. According to reports, OpenAI supported the signing, but xAI's Elon Musk and Meta's Mark Zuckerberg helped derail it — a claim Musk has since denied on X. A key point of friction was the timeline. The government sought up to 90 days to evaluate models pre-release, while AI labs pushed for a much shorter 14-day window. The consensus among DC watchers is that this highlights a deep tension inside the administration itself, between security-focused agencies pushing for governance now and others favouring a lighter regulatory touch to maintain momentum.
Sailfish's take
We're reading this as a temporary win for speed, not a final verdict on safety. The underlying pressure for government oversight hasn't vanished — it has just been delayed by a mix of logistics and lobbying. For builders, this means the regulatory environment stays permissive for a little while longer, which is useful for shipping quickly. But we wouldn't bet the farm on it lasting. The debate over a 90-day versus a 14-day review window shows the gap is wide. We think some form of pre-deployment testing is inevitable. The useful question isn't if regulation is coming, but what form it will take. For now, ship fast.