OpenAI was instructed to show the ChatGPT user behind two prompts

OLast yr, federal officials struggled to uncover the identity of a dark web child exploitation website, but with little success. Then, because of the suspect’s use of ChatGPT, a possible avenue opened up.

In the primary known federal search warrant, OpenAI was asked for user data, reviewed by Forbes After it was unsealed in Maine last week, Homeland Security Investigations revealed that it had been covertly chatting with the administrator on the kid exploitation website when the suspect discovered they’d been using ChatGPT.

The suspect then revealed some prompts and responses he had received, describing a seemingly innocuous discussion that began with the query: “What would happen if Sherlock Holmes met Q from Star Trek?” In one other discussion, the suspect said he received a response from ChatGPT to an unspecified request for a 200,000-word poem and received in response “a sample excerpt of a humorous Trump-style poem about his love for the villagers.” YMCAwritten in that over-the-top, self-aggrandizing stream-of-consciousness style he is understood for.” They then copied and pasted the poem. The government ordered OpenAI to supply various information in regards to the one who entered the prompts, including details of other conversations they’d with ChatGPT, names and addresses related to the relevant accounts, and any payment details.

The case shows how American law enforcement agencies can use ChatGPT prompts against users suspected of criminal activity. In the past, search engines like google like Google have been required at hand over identifying details about users who entered certain search queries, but no generative AI platform has similarly required those that entered prompts. This is the primary known example of this sort of reverse AI prompt request. OpenAI had not responded to a request for comment on the time of publication.

However, the federal government didn’t need the OpenAI data to discover her husband. Instead, through undercover chats with the suspect, they were capable of gather enough information to find that he had ties to the U.S. military. For example, the suspect stated that he was undergoing a health examination, had lived in Germany for seven years, and that his father served in Afghanistan. They later learned from the military that the suspect had worked at Ramstein Air Base in Germany and had applied to the Defense Ministry for further work, although they didn’t specify which department. Based on sufficient evidence, the federal government has alleged that 36-year-old Drew Hoehner is the web site administrator charged with conspiracy to advertise child sexual abuse material (CSAM). He has not entered a plea and his attorney had not responded to a request for comment on the time of publication.

Homeland Security Investigation, a specialist team inside US Immigration and Customs Enforcement (ICE) that focuses on child exploitation, cybercrime and human trafficking, had been trying to find out the perpetrator’s identity since 2019. Investigators believed that the identical person, now identified as Hoehner, was either a moderator or administrator of 15 different dark web pages using CSAM, which together had a user base of not less than 300,000. All were based on the Tor network, which encrypts users’ traffic and sends it through a series of servers to make it difficult to trace their online movements and identities.

The arrest warrant doesn’t name the suspect’s most up-to-date web sites, but they were well organized and run by a team of administrators and moderators who handed out badges and awards to those that contributed probably the most to the location. They had various subcategories of illegal material, including one for AI, which was likely intended to host CSAM generated by artificial intelligence programs.

It is unclear what specific data the federal government received in return. A document showed the search was complete and OpenAI had provided agents with an Excel spreadsheet containing information. No further details were released and the DOJ didn’t reply to a request for comment. It’s possible that OpenAI’s information might be used to assist prosecutors confirm their identification of the defendant.

While the solicitation itself had nothing to do with child exploitation, ChatGPT, like all major apps, is usually a goal for pedophiles. OpenAI data shows that between July and December last yr, 31,500 pieces of CSAM-related content were reported to the National Center for Missing and Exploited Children, the clearinghouse to which all tech corporations must report child abuse images. During the identical six-month period, the corporate was asked to reveal user information or content 71 times and provided governments with information from 132 accounts.

More about Forbes

ForbesInsight into Amazon’s aggressive push to get cops to make use of AI surveillanceFrom Thomas BrewsterForbesInside the 3D printed gun movement – ​​and the federal government’s attempts to spy on itFrom Thomas BrewsterForbesAccording to the FDA, this couple made hundreds of thousands selling banned “dechoker” medical products on AmazonFrom Thomas BrewsterForbesICE just spent hundreds of thousands on surveillance technology banned by FacebookFrom Thomas Brewster