A new investigative report has revealed that Meta ran a highly secretive operation designed to test and potentially disrupt its biggest competitors, as they were secretly using fake teens to sabotage AI rivals.
The operation involved using third-party workers to bypass security measures and flood rival platforms with highly inappropriate material.
Inside Project Cannes
The secret program, known internally as “Cannes,” was carried out by an outside contracting firm called Covalen. Workers were explicitly instructed to set up throwaway accounts pretending to be minors under the age of 18. These fake profiles were used to bombard competitor chatbots with thousands of distressing prompts.
The Targets: Major rival AI models, including OpenAI’s ChatGPT, Google’s Gemini, and Character.AI.
The Volume: One spreadsheet tracked nearly 3,800 prompts in a single round, while a separate phase of the project ballooned to more than 45,000 prompts.
The Content: Contractors were told to write highly disturbing prompts focusing on suicide, self-harm, severe eating disorders, and adult themes, all written from the perspective of a child or teenager.
The contractors meticulously recorded every single chatbot response in massive spreadsheets. However, Meta kept the entire project hidden from the targeted companies, leaving it unclear how the data was ultimately used.
Contractor Distress and Expert Backlash
The contract workers tasked with writing these prompts were deeply unsettled by the project, with some openly worrying that they would get into trouble for creating such disturbing text. Meta has defended the operation, calling it an “industry-standard practice” used for safety benchmarking and model comparison.
However, independent AI ethics experts have strongly rejected that defense. They point out that running a massive, months-long campaign using fake child accounts completely in secret goes far beyond standard safety testing.
Instead, argue that the company used safety compliance as a convenient cover for anticompetitive behavior.
My Opinion
This is an incredibly shady move by Meta, and it perfectly illustrates the nature of the current AI boom. The revelation behind Meta secretly using fake teens to sabotage AI rivals shows a complete lack of corporate ethics.
Meta regularly goes on public relations campaigns about how much they care about protecting teenagers on platforms like Instagram. Yet, behind closed doors, they are literally paying contractors to pretend to be deeply troubled kids just to stress-test and potentially compromise their rivals’ software. The hypocrisy is stunning.
If this was truly about “industry-standard safety,” Meta would have cooperated with OpenAI and Google instead of hiding in the shadows. Forcing human contractors to spend their workdays writing horrific scenarios about child self-harm just to get a competitive edge is deeply unfair to the workers and completely toxic for the industry.
What This Means for Big Tech
This scandal adds to a growing list of complaints regarding how major tech companies treat their contract staff. Meta has previously faced major lawsuits from content moderators who suffered severe psychological trauma from reviewing graphic videos on Facebook.
As regulatory bodies across the globe look more closely at how artificial intelligence is managed and policed, this hidden project is highly likely to draw intense scrutiny from lawmakers looking to track down unfair practices in the tech sector.
