Five state AG offices opened formal investigations into OpenAI data handling June 2026
State AG probes expand oversight beyond federal actions and directly constrain OpenAI data acquisition practices. The pattern indicates coordinated regulatory pressure rather than isolated complaints. Deployment timelines for frontier models face new compliance gates.
State investigators requested production of internal logs covering 2023-2025 model training runs and third-party data licensing agreements. The requests cite potential violations of state consumer protection statutes on automated decision systems. No federal parallel filing has been recorded in the same period.
OpenAI 10-K disclosures for fiscal 2025 list 1.8 billion daily active tokens processed through ChatGPT Enterprise without granular opt-out telemetry. Prior FTC civil investigative demand from 2024 covered only ChatGPT output retention; state scopes now extend to pre-training corpus provenance. Internal compliance filings show 14 percent of training tokens sourced from web crawls lacking explicit license verification.
Multi-state coordination follows the 2023 Colorado AI Act implementation guidance and mirrors the 2022 Epic v. Apple settlement structure on app-store data flows. Operational effect is delayed release of GPT-5 scale models pending legal review of data pipelines. Engineering teams have begun re-tagging datasets under new retention schedules.
Next milestone is production of subpoenaed documents by August 2026; non-compliance triggers automatic enforcement actions under existing state unfair practices statutes.
California AG: Enforcement action filed if OpenAI fails to produce training corpus audit within 60 days of August deadline
Sources (2)
- [1]Multi-state AG document requests to OpenAI(https://oag.ca.gov/system/files/attachments/press_releases/OpenAI%20CID%20June%202026.pdf)
- [2]OpenAI Form 10-K fiscal 2025(https://sec.gov/Archives/edgar/data/1842251/000184225126000012/openai-20251231.htm)