Meta hired contractors to pose as teenagers and test ChatGPT

According to WIRED, Meta paid hundreds of contract workers through contractor Covalen to create fake accounts posing as teenagers under 18 and test competitors' chatbots. The project, code-named Cannes, remained active at least until April 21, 2026.

How the Cannes project worked

Covalen's contractors signed up for accounts listing an age under 18, then sent text prompts and images to chatbots from other companies. They logged the systems' responses by hand in spreadsheets. In effect, this was systematic data collection on how rival models respond to users who present themselves as minors.

• The project's code name was Cannes
• Contractor Covalen managed the workers
• Accounts were created in the name of users under 18
• Chatbot responses were recorded manually in spreadsheets
• The project was active at least until April 21, 2026

Which services were targeted

WIRED names OpenAI's ChatGPT among the products that Cannes contractors tested. The publication does not disclose the full list of other services the project focused on, nor whether testing was limited to text prompts or included more complex conversation scenarios.

Why this raises questions

Using fake accounts of minors to test other companies' products raises two questions. First, it likely violates competitors' terms of service: most chatbots prohibit automated data collection and creating accounts with false age information. Second, the project's mechanics show that big tech companies care not only about the safety of their own underage users but also about how competitors handle that audience — and they chose to find out covertly rather than through open audits or partnerships with independent researchers.

Protecting underage chatbot users remains one of the most sensitive issues for the generative AI industry: regulators and parents' groups regularly point to the risks of harmful content and manipulative conversations. The Cannes project adds a new element to this picture — competitive intelligence gathered while posing as a vulnerable audience.

Stories like this also shed light on a broader practice among tech companies: competitive intelligence in AI has long since moved beyond public benchmarks and open model comparisons. Companies hire third-party agencies to systematically monitor how other services respond — and they do it not under their own name but through contractors like Covalen, which makes it harder to attribute such activity directly. If WIRED's reporting on Cannes is confirmed, it will become one of the most concrete public examples of this practice carried out by Meta specifically.

What this means

The Cannes story shows that covert testing of rival products while posing as teenage users has become part of the competition in the chatbot market, alongside open benchmarks and public comparisons. For the industry, it's another argument for stricter user age verification and tighter controls on automated queries to chatbots.

Frequently asked questions

What is the Cannes project?

It's the internal name of an initiative in which contract workers from Covalen, a contractor working for Meta, created fake accounts for users under 18 and sent prompts and images to rival chatbots, logging the responses in spreadsheets. The project remained active at least until April 21, 2026, WIRED reports.

Who is Covalen, and what does Meta have to do with it?

Covalen is a contractor that, according to WIRED, managed hundreds of contract workers on the Cannes project. Those workers created the fake teen accounts and tested rival chatbots, including OpenAI's ChatGPT.