What happened
Introduced in the final week of September 2026 (reported 29 September), the AI Systems Transparency Act (sponsored by Sens. Britt, Coons, Lankford and Schatz) would require AI companies to publicly disclose data their models collect from users, child/adult safeguards, guardrails against 'going rogue' or exploitation, model-card-style documentation, and to update disclosures with each model release; the FTC would enforce, covering closed- and open-source models above a 1M-monthly-user threshold or with minor accounts.
Why it matters
ASTA is one of the most developed transparency bills in the current crop and would give the FTC direct jurisdiction over model disclosure and safety documentation — a concrete mechanism to make AI labs' safety claims verifiable and enforceable, complementing the FTC's active investigation.
Action needed
AI companies with US-facing models should prepare public disclosure programmes covering training data, user-data collection, child-safety, and loss-of-control safeguard descriptions if ASTA advances.