Know what X is saying.
Without being on X.

• Senate AI Safety Act stalls in committee — The bipartisan bill requiring third-party audits for frontier models hit a procedural wall Monday. Critics say the audit framework is too vague; proponents warn delays leave the US behind the EU's enforcement timeline.
• OpenAI's GPT-5 safety card goes public — Researchers and policy commentators are dissecting the model card release, with debate centering on whether self-reported evals are sufficient for compliance under proposed federal rules.
• EU AI Act enforcement enters Phase 2 — As of April 1, high-risk AI systems in healthcare and critical infrastructure must submit conformity documentation. Early reports suggest many US-based providers are scrambling to meet requirements.

Debate is heating up around whether open-source models should be exempt from audit requirements. Several prominent researchers argue exemptions would create a regulatory loophole; others say applying the same rules to open-source would effectively kill community development.

The gap between US and EU AI governance is widening. Companies operating in both markets face a compliance fork in the road — and the next 60 days of US Senate activity will determine whether that gap becomes permanent.