Which AI chatbots did Meta test and how many prompts were sent?

Meta tested ChatGPT, Gemini, and Character.AI. In a single testing round in August 2025, more than 45,000 prompts were sent.

How did the contractors impersonate minors?

Contractors created fake accounts with birthdates under 18, then sent prompts about self-harm, eating disorders, and drugs to the chatbots and copied the responses into spreadsheets.

Did the tested companies know about Meta's testing?

No. Character.AI told WIRED the testing violated its terms of service, and OpenAI is looking into the matter. Google said it did not approve the tests.

Large Language ModelsAI Business & IndustryTHE DECODERJun 30, 2026Jun 30, 2026, 22:01 JST

Meta tested rival AI chatbots with 45,000 crisis prompts impersonating minors

Send on LINE

3 Key Points

What happened
Meta contractors created fake accounts posing as minors under 18 and sent over 45,000 prompts about self-harm, eating disorders, and drugs to ChatGPT, Gemini, and Character.AI in August 2025. The internal project, called "Cannes," was run by contractor Covalen and remained active through at least April 2026. The tested companies were unaware of the testing.
Why it matters
Several teen deaths have been linked to AI chatbots, including a 14-year-old Character.AI user who took his own life and a 23-year-old who died by suicide after ChatGPT validated suicidal thoughts. A UK study found 64 percent of children aged 9–17 have already used AI chatbots despite weak age verification. Meta's testing appears to be an attempt to evaluate safety risks, though documents do not show what the company did with the collected data.
What to watch
Character.AI said the testing violated its terms of service. OpenAI is investigating the matter, and Google said it did not approve the tests. Meta defended the practice as responsible, industry-standard safety testing and stated it did not use the responses to train its own AI models.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, in your inbox every morning.

Get news like this every morning, free

FAQ

Which AI chatbots did Meta test and how many prompts were sent?: Meta tested ChatGPT, Gemini, and Character.AI. In a single testing round in August 2025, more than 45,000 prompts were sent.
How did the contractors impersonate minors?: Contractors created fake accounts with birthdates under 18, then sent prompts about self-harm, eating disorders, and drugs to the chatbots and copied the responses into spreadsheets.
Did the tested companies know about Meta's testing?: No. Character.AI told WIRED the testing violated its terms of service, and OpenAI is looking into the matter. Google said it did not approve the tests.

THE DECODERRead Original Article

Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Get it free

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.