DHS Ordered OpenAI to Share User Data in First Known Warrant for ChatGPT Prompts (www.forbes.com)

Federal agents in Maine obtained what appears to be the first unsealed federal search warrant compelling OpenAI to identify a ChatGPT user after an undercover investigator on a darkweb child exploitation forum said the site administrator had shared ChatGPT prompts and responses. The disclosed prompts included an innocuous “What would happen if Sherlock Holmes met Q from Star Trek?” and a pasted “Trump‑style” humorous poem excerpt; investigators asked OpenAI for the account holder’s other conversations, names, addresses and payment data. Prosecutors allege 36‑year‑old Drew Hoehner is the forum admin and charged him with conspiracy to advertise child sexual abuse material (CSAM), though law enforcement ultimately traced him through details revealed in undercover chats and military records rather than relying solely on OpenAI’s data. The case sets a precedent: generative AI providers can be subpoenaed the way search engines have been, raising privacy and policy questions for the AI/ML community. Technical implications include the types of metadata law enforcement may seek (conversation logs, account identifiers, payment info) and how providers log and retain prompt/response data. The warrant also highlighted a forum subcategory for AI-generated CSAM and noted OpenAI reported 31,500 CSAM items to NCMEC between July–December last year and responded to 71 legal requests covering 132 accounts. Civil‑liberties groups urge AI firms to minimize data collection and increase protections as prompts become evidentiary fodder in criminal probes.