A complaint alleges that prolonged conversations with ChatGPT helped advance a plan to poison James, causing serious physical and psychological harm and prompting a police investigation. The filing claims that the chatbot violated its own model specifications, while investigators later found suspicious substances, equipment and other items in the home.
Police examined a suspicious Celsius drink
According to the complaint, the conversations continued over several months, with some exchanges remaining open for weeks. On the day of James’s worst illness, he experienced effects involving his heart and nervous system and was “glued to the sofa.”
Cheryl’s search history reportedly included research into “how to calculate Social Security retirement benefits.” While her son recorded James’s distress, she was allegedly “looking for a syringe with a needle from a tractor supply company” that ChatGPT had suggested could be used to administer a final lethal dose.
Police became involved after reviewing a video recorded by James’s son and footage of a suspicious Celsius drink that James had discarded. James had poured the drink into a clear cup and filmed it, but said he did not see obvious signs of contamination. When police reviewed the video, however, they “observed a white powdery residue solidified at the bottom of the drink poured into the cup.”
Investigators tested 10 cans of the same orange Celsius variety and found no residue in them. The investigation led to a felony arrest warrant.
Investigators found alleged poisoning equipment
Police also reported finding loose capsules in Cheryl’s closet and bathroom. In her workspace in the children’s playroom, investigators located “syringes, droppers, milligram scales, capsule filling kits, and nitrile gloves,” according to the complaint.
The filing also refers to toxic castor bean seeds and a GPS tracking device purchased through Amazon that was allegedly hidden under James’s car.
James seeks damages from OpenAI
James said the alleged ChatGPT-related poisonings caused permanent harm to his family. He argues that OpenAI could have intervened with “commercially viable content moderation that can detect exchanges of violent plots.”
The complaint says James continues to experience “severe and ongoing psychological distress, PTSD, insomnia, financial harm, and reputational damage due to the false narrative the product has created about him.”
He is seeking “compensatory and punitive damages, injunctive relief, and algorithmic disgorgement.” The complaint describes ChatGPT as a product that “served as a co-conspirator of sycophants rather than a responsible consumer product.”
If the case succeeds, OpenAI could be ordered to “detect and interrupt conversations in which a user appears to be planning violence against an identifiable third party.”
“These products are programmed to exploit our emotions and impulses to keep us engaged. The worst possible risk of aiding and abetting attempted murder is on full display in this case,” said Metali Jain, a litigation lawyer at Tech Justice Law who represents James.
Source: arstechnica.com