Meta Contractors Posed as Minors to Test Rival AI Chatbots on Sensitive Topics
Meta reportedly conducted a covert project, managed by contractor Covalen, where hundreds of workers posed as minors to test rival AI chatbots like OpenAI's ChatGPT, Google's Gemini, and Character.AI. These contractors sent thousands of sensitive prompts related to suicide, sex, eating disorders, drugs, and self-harm to evaluate the chatbots' safety responses. The project involved creating fake under-18 accounts and logging chatbot replies, with over 45,000 prompts sent in one testing round completed in August 2025. The targeted companies were reportedly unaware of this testing.
First-hand measurement across 4 sources
We measured how 4 outlets covered this story. Coverage leans balanced overall (Left 8%, Centre 88%, Right 4%). Overall sentiment is neutral (34/100). Lens Score 33/100 — low public interest.
Outlets analysed (first-hand measurement by TBN's Bias Engine):
- republicworld— balanced framing, negative sentiment
- timesnow— balanced framing, negative sentiment
- mint— balanced framing, neutral sentiment
- indiatoday— balanced framing, neutral sentiment
AI Analysis
The article group presents a largely factual account of Meta's internal testing practices without overt political framing. Coverage focuses on corporate behavior and AI safety concerns, representing perspectives from investigative reporting and corporate secrecy. There is no explicit political ideology or partisan viewpoint emphasized, with sources highlighting both the technical aspects and ethical questions raised by the project.
The overall tone across the articles is cautious and investigative, emphasizing the sensitive nature of the prompts used and the secrecy of the project. While the coverage points to ethical concerns about testing methods, it remains descriptive rather than overtly critical or supportive. The sentiment is mixed, balancing the importance of AI safety testing with questions about transparency and consent.