DHS Asks OpenAI To Share Information on ChatGPT Prompts Used By Users
\n \t
\n
<\/BR>
\n
\n \t
\n
<\/BR><\/p>\n
The Department of Homeland Security (DHS) has issued the first known federal search warrant compelling OpenAI to disclose user data tied to ChatGPT prompts.<\/p>\n
The warrant, unsealed last week in Maine and reviewed by cybersecurity outlets, stems from a year-long probe into a dark web site distributing child sexual abuse material (CSAM). <\/p>\n
Federal agents, operating undercover, hit a breakthrough when the site\u2019s administrator casually revealed their use of ChatGPT during online chats.<\/p>\n
The suspect, chatting with investigators posing as site users, shared snippets of interactions with the AI tool. <\/p>\n
One prompt explored a whimsical crossover: \u201cWhat would happen if Sherlock Holmes met Q from Star Trek?\u201d Another involved requesting a 200,000-word poem, to which ChatGPT responded with a sample a bombastic, self-aggrandizing ode in the style of President Donald Trump praising the Village People\u2019s \u201cY.M.C.A.\u201d <\/p>\n
The administrator even pasted the full excerpt, unwittingly providing investigators with a digital trail.<\/p>\n
The warrant seen by Forbes<\/a> directed OpenAI to surrender extensive records on the account behind these prompts.<\/p>\n This included full transcripts of other ChatGPT conversations<\/a>, associated names, email addresses, IP logs, and payment details linked to subscriptions.<\/p>\n Such requests echo past law enforcement demands on search engines like Google for user queries, but mark the debut of generative AI platforms.<\/p>\n Experts warn this could set a precedent, allowing agencies to reverse-engineer criminal intent from seemingly benign AI interactions.<\/p>\n OpenAI has not publicly commented on the warrant or its compliance. Privacy advocates, however, raise alarms about the chilling effect on AI users, arguing that innocuous prompts could now flag everyday citizens in broad surveillance nets.<\/p>\n Ironically, the OpenAI data proved unnecessary for identification. Undercover exchanges revealed the suspect\u2019s ties to the U.S. military: mentions of health assessments, seven years in Germany at Ramstein Air Force Base, and his father\u2019s service in Afghanistan.<\/p>\n Cross-referencing with Department of Defense records confirmed 36-year-old Drew Hoehner as the administrator. <\/p>\n Hoehner, who had applied for further DoD roles, faces one count of conspiracy to advertise CSAM. He has yet to enter a plea, and his attorney did not respond to inquiries.<\/p>\n This case underscores evolving tactics in cybercrime investigations, where AI tools become unwitting informants. As generative platforms proliferate, law enforcement\u2019s reach into user creativity grows, potentially eroding trust in tools like ChatGPT<\/a>.<\/p>\n For now, it highlights how a single shared prompt can unravel hidden networks, blending pop culture queries with serious allegations.<\/p>\n Follow us on Google News<\/a>, LinkedIn<\/a>, and X<\/a> for daily cybersecurity updates. Contact us<\/a> to feature your stories.<\/strong><\/p>\n The post DHS Asks OpenAI To Share Information on ChatGPT Prompts Used By Users<\/a> appeared first on Cyber Security News<\/a>.<\/p>\n<\/div>\n