OpenAI Faces DHS Request to Disclose User’s ChatGPT Prompts in Investigation

Cybersecurity: Federal Warrant for AI Data

Recently, a federal investigation has resulted in a warrant seeking access to AI-generated chat logs. Homeland Security Investigations (HSI) in Maine requested user data from OpenAI, marking a notable instance where AI data is sought as evidence.

The investigation involved undercover agents communicating with a suspect operating a child exploitation site. The suspect disclosed using ChatGPT, sharing some responses generated by the AI. The prompts included various topics, ranging from fictional scenarios to creative writing requests.

HSI's request to OpenAI encompassed all available user data, including conversations, account names, contact details, and payment records. This mirrors previous data requests to search engines but is unprecedented in the context of generative AI.

Despite the request, investigators identified the suspect without relying on AI records. The suspect, Drew Hoehner, was identified through personal details disclosed during ongoing communications. Hoehner, with military connections, was charged with conspiracy to advertise child sexual abuse material (CSAM).

The investigation revealed connections to multiple CSAM sites on the Tor network, involving potentially 300,000 users. Although the site's specific details remain confidential, it reportedly included sections for AI-generated illegal content.

The outcome of OpenAI's response to the warrant is not fully known, though officials received a spreadsheet of user data that may aid in corroborating charges. OpenAI reported handling thousands of CSAM-related instances and responded to numerous government data requests.

The warrant focused on specific prompts from one user, highlighting a trend of law enforcement seeking AI as evidence. This raises concerns about data privacy and the extent of data collection by AI companies.

Based on reporting by GBHackers.