Meta Tested Competitor Chatbots With Simulated Minor Crisis Prompts

Meta employed hundreds of contractors to impersonate minors and submit prompts involving suicide, sex, and drugs to chatbots developed by OpenAI, Google, and Character.AI. In one round alone, more than 45,000 such prompts were delivered without the knowledge of the targeted companies. The undisclosed testing raises questions about industry practices for evaluating AI responses to sensitive scenarios.