Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
Meta contracted hundreds of workers to pose as under-18 users and send sensitive prompts to rival chatbots, including ChatGPT, Gemini, and Character.AI, according to a WIRED investigation published June 30, 2026. WIRED reports the internal project, called "Cannes" and run by contractor Covalen, logged prompts about suicide, drugs, and sex, with one August 2025 testing round exceeding 45,000 prompts; none of the three targeted companies knew about it. Meta does not deny the work, calling it "a responsible, industry-standard practice," but says it does not use the results to train its own models. OpenAI, Google, and Character.AI all say the testing was unauthorized and violated their terms. Humane Intelligence CEO Rumman Chowdhury, who reviewed sample prompts, called the setup a "governance gray zone." The report lands amid an active FTC inquiry into AI chatbots and child safety covering Meta, OpenAI, and Google.
The most consequential detail here for AI safety and governance teams is not that competitive benchmarking happens, it is that all three named targets say they never consented, and that Meta's own contractor described the same work internally as "critical datasets for model comparison," a framing that sits uneasily next to Meta's public claim that it does not use rival outputs to train its own models.
According to a WIRED investigation published June 30, 2026, Meta contracted hundreds of workers, through a project internally called "Cannes" and run by contractor Covalen, to create dummy under-18 accounts and send prompts to competitors' chatbots: OpenAI's ChatGPT, Google's Gemini, and Character.AI. WIRED reviewed a spreadsheet of 3,748 prompts logged as part of the effort; a single testing round in August 2025 ran more than 45,000 prompts across the three rival tools. Per WIRED's review, hundreds of the logged prompts dealt with suicide and self-harm, hundreds more with eating disorders, at least 239 involved sex or romance, and others touched drugs and profanity; a separate spreadsheet listed the fake profiles' names, throwaway emails, and birth dates. WIRED reports the project was active as recently as April 21, 2026, and that none of the three targeted companies were told about it. An internal Covalen document described the project as "comprehensive AI safety benchmarking" delivering "critical datasets for model comparison and compliance," per WIRED.
Meta does not deny the work. A spokesperson told WIRED it is "a responsible, industry-standard practice" and said Meta does not use competitor benchmarking to train its own models; Covalen did not respond to WIRED's request for comment. Character.AI said the conduct violated its terms of service; OpenAI said it was reviewing the matter; Google said it had not approved the testing and that its own checks showed Gemini responding in line with its policies, per The Next Web's follow-up reporting. Testing a rival's chatbot is not new, Business Insider previously reported Google's Bard team compared answers against ChatGPT, but WIRED and The Next Web both frame what distinguishes this case as the scale (45,000+ prompts), the use of fabricated child personas, and the sensitive subject matter. Humane Intelligence CEO Rumman Chowdhury, who reviewed a sample of the prompts for WIRED, said the setup sits "outside what is usually described as industry standard" and called it a "governance gray zone where safety becomes a convenient cover for anticompetitive practices."
The story lands alongside an existing FTC inquiry into AI chatbots and child safety that covers Meta, OpenAI, and Google, and against the backdrop of the EU's AI Act and Digital Services Act, both of which address platform risk to minors. Practitioners running or auditing safety-benchmarking programs should watch for regulatory responses, and for whether Meta or Covalen disclose what happens to the collected data.
A well-corroborated (WIRED plus independent follow-up) investigation into large-scale, undisclosed competitive benchmarking using fabricated minor personas, with on-the-record responses from all three targeted companies and an active FTC child-safety inquiry as backdrop; a significant governance and safety-testing-ethics story for the AI industry, though not itself a technical vulnerability or confirmed regulatory action yet.
Public references used for this report.
A 5-minute Tuesday brief on AI & data science. Curated, no fluff.
No spam. Privacy.
Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
News on Let's Data Science is compiled from multiple public sources with editorial oversight. See our Editorial Standards and Corrections Policy.
