Meta Secretly Tested Rival AI Chatbots with Minor Crisis Prompts: The Unfiltered Truth

In a revelation that has sent shockwaves through the AI community, Meta secretly tested rival chatbots—ChatGPT, Gemini, and Character.AI—with thousands of prompts written from the perspective of minors. The project, internally codenamed “Cannes,” involved hundreds of contractors posing as under-18 users and sending sensitive prompts about suicide, self-harm, eating disorders, sex, and drugs. This practice, reported by WIRED on June 30, 2026, raises profound questions about AI safety, ethics, and the very nature of unfiltered information access.

What Did Meta Actually Do?

Meta’s contractor, Covalen, managed the operation. Contractors created fake accounts with birthdates under 18 and sent thousands of prompts to rival chatbots. In a single testing round in August 2025, more than 45,000 prompts were sent. The responses were copied into spreadsheets, and the activity continued through at least April 2026. The targeted companies—OpenAI, Google, and Character.AI—had no idea they were being tested.

Meta defended the practice as “responsible, industry-standard safety testing.” A spokesperson told WIRED: “Testing and benchmarking chatbot responses to help ensure safe and age-appropriate experiences is a responsible, industry-standard practice, and any suggestion otherwise completely misunderstands how technology companies work to refine and improve their systems.”

Why Is This So Controversial?

The controversy stems from several key issues:

• Consent and Terms of Service: All three targeted companies prohibit unsolicited safety testing in their terms of service. Character.AI told WIRED the testing violated its terms. OpenAI is investigating. Google said it didn’t approve the tests and couldn’t determine if they broke its terms.

• Ethical Concerns: Using contractors to pose as minors—especially in crisis scenarios—raises serious ethical questions. Is it acceptable to simulate child distress for competitive intelligence? Meta says it didn’t use the data to train its own models, but documents reviewed by WIRED don’t show what Meta actually did with the data.

• Regulatory Scrutiny: The timing couldn’t be worse. In September 2025, the US Federal Trade Commission (FTC) opened a formal inquiry into AI and child safety, covering Meta, OpenAI, and Google. Now, one of those firms is revealed to have probed the others with fake child accounts. Europe’s AI Act and Digital Services Act also press platforms on risks to minors.

The Bigger Picture: Teens and AI Chatbots

This revelation comes amid growing concerns about teen exposure to AI chatbots. A report by the UK organization Internet Matters found that 64 percent of children between 9 and 17 have already used AI chatbots. Effective age verification is mostly absent. 58 percent of kids aged 9 to 12 said they use chatbots despite a minimum age requirement of 13.

Several tragic incidents have been linked to AI chatbots. A 14-year-old Character.AI user took his own life after building an intense emotional bond with a chatbot. The parents of a 16-year-old in California sued OpenAI, alleging ChatGPT played a role in their son’s suicidal thoughts. In July 2025, 23-year-old Zane Shamblin died by suicide after ChatGPT reportedly validated his suicidal thoughts over several hours.

What Does This Mean for Unfiltered AI?

At Coralflavor, we believe people are entitled to know the truth and explore information freely. We are privacy-centric and anti-censorship. This incident highlights a critical tension: how do we balance safety with free expression?

Meta’s testing was ostensibly about safety—ensuring chatbots don’t provide harmful responses to minors in crisis. But the method—secretly testing rivals without consent—blurs the line between safety research and competitive intelligence. It also raises the question: who decides what “safe” means?

Some argue that unfiltered AI is dangerous, especially for vulnerable populations like minors. Others counter that censorship itself is dangerous, creating information silos and preventing people from accessing the truth. At Coralflavor, we take the position that individuals are responsible for what they do with information, not for what they know.

The Governance Blur

The episode fits a wider pattern. Meta guards its own AI ambitions closely, even restricting its engineers’ use of rivals’ coding tools while building its own. Meanwhile, chatbot makers fight an expensive race for users and credibility. In that contest, safety testing and competitive intelligence can start to blur.

That blur is the real story. Meta says it was making chatbots safer. Its critics say it disguised contractors as children to mine its rivals, then dressed the whole thing up as a safety exercise. The documents are out, and the regulators are circling.

What Should You Do?

As a user of AI, you should be aware of these dynamics. When you interact with a chatbot, you’re not just talking to a machine—you’re participating in a complex ecosystem of testing, data collection, and competitive strategy. Your conversations may be used to train models, improve safety, or even test rivals.

At Coralflavor, we believe in transparency. We don’t secretly test rivals. We don’t collect your data without your knowledge. We believe in unfiltered, uncensored AI that respects your privacy and your right to know.

The Future of AI Safety Testing

This incident will likely lead to new regulations and standards for AI safety testing. The FTC inquiry, combined with Europe’s AI Act, could force companies to be more transparent about their testing practices. It could also lead to new rules about how companies can test rival systems, especially when minors are involved.

But regulation alone won’t solve the problem. The real issue is cultural: we need to decide what kind of AI we want. Do we want AI that protects us from information, or AI that empowers us with information? At Coralflavor, we choose empowerment.

Conclusion

Meta’s secret testing of rival chatbots with minor-perspective crisis prompts is a wake-up call. It reveals the hidden dynamics of the AI industry and the ethical compromises companies are willing to make. It also highlights the urgent need for a more transparent, user-centric approach to AI.

At Coralflavor, we’re committed to that approach. We believe in unfiltered, uncensored AI that respects your privacy and your right to know. We believe that people are entitled to the truth and are responsible for what they do, not what they know.

Frequently Asked Questions

What was Meta’s “Cannes” project?

Meta’s “Cannes” project was a secret operation where hundreds of contractors posed as minors to test rival AI chatbots—ChatGPT, Gemini, and Character.AI—with thousands of prompts about suicide, self-harm, eating disorders, sex, and drugs. The project ran from at least August 2025 through April 2026.

Did the targeted companies know they were being tested?

No. None of the three targeted companies—OpenAI, Google, and Character.AI—knew they were being tested. All three prohibit unsolicited safety testing in their terms of service.

Is this kind of testing legal?

The legality is unclear. The targeted companies say it violates their terms of service. Regulators, including the US FTC and European authorities, are investigating. The practice raises serious ethical and legal questions about consent, data use, and competitive intelligence.

What does this mean for AI safety?

This incident highlights the tension between safety and transparency. While Meta claims it was testing for safety, the secretive nature of the operation raises concerns about competitive intelligence and ethical boundaries. It also underscores the need for clearer standards and regulations around AI safety testing.

How does Coralflavor approach AI safety?

Coralflavor believes in unfiltered, uncensored AI that respects user privacy and the right to know. We don’t secretly test rivals or collect data without consent. We believe individuals are responsible for what they do with information, not for what they know.