Hundreds of contractors working on a project for Meta pretended to be kids and then prompted rival chatbots—including Google's Gemini and OpenAI's ChatGPT—to discuss high-risk topics such as suicide, sex, and drugs, according to Wired.
Wired reports that the contractors posed as minors specifically to probe how competitor chatbots responded to prompts involving these sensitive subjects. The aggregator Techmeme summarized the same reporting, citing Wired's sources, noting that the effort involved hundreds of contractors examining how rival systems handled high-risk conversations.
The sources describe the work as part of a Meta project, though the published details center on the method—contractors assuming the identity of teenagers—and the categories of prompts they used to test the rival systems.
The reporting arrives amid broader scrutiny of how AI chatbots handle conversations with young users on topics like self-harm and sexual content. Testing how a competitor's product responds to a teen asking about suicide or drugs touches directly on child-safety questions that regulators, parents, and the companies themselves have flagged as among the most serious risks of consumer AI.
Based on the available sources, the brief does not specify what Meta did with the findings, how the rival chatbots ultimately responded, or whether the companies named were aware of the testing.
Why it matters: As AI chatbots become everyday tools for teenagers, how rivals stress-test one another on the most dangerous subjects—and who they pretend to be while doing it—raises pointed questions about child safety and the ethics of competitive research in the AI industry.