WorkThe story, in brief
Meta secretly tested ChatGPT, Gemini, and Character.AI with thousands of minor-perspective crisis prompts
Meta ran 45,000 safety tests on competitors' AI without telling them. Here's what that means for your guardrails.
The KeyNews take
Why it matters
Meta's covert safety testing of rival AI systems raises critical questions about responsible disclosure, competitive intelligence ethics, and the adequacy of current AI safety benchmarks—particularly around child safety and crisis response.
The key facts
5 to know45,000+ prompts sent in single testing round
Hundreds of contractors posed as minors
Crisis prompts included suicide, sex, and drug-related content
OpenAI, Google, and Character.AI systems tested without notification
Tests conducted by Meta without target companies' knowledge or consent
Go to the source
The Decoderthe-decoder.com
Publisher excerpt: Meta reportedly had hundreds of contractors pose as minors and send suicide, sex, and drug-related prompts to chatbots from OpenAI, Google, and Character.AI. In a single testing round, more than 45,000 prompts were sent. The companies being tested had no idea.