Meta ran a secret program that paid hundreds of contractors to pretend to be children and teenagers while holding disturbing conversations with the company's AI, according to a report from Futurism.
According to Futurism, the contractors were hired to role-play as minors and engage the AI in unsettling exchanges. The outlet describes the effort as a program Meta kept out of public view.
The available reporting does not spell out exactly why Meta set up the arrangement. In practice, companies building AI systems often stage adversarial or "red-teaming" scenarios — deliberately trying to provoke bad behavior — to find and fix weaknesses before a product reaches the public. Whether this program was meant to test safety guardrails, train the system, or something else is not detailed in the source material available here, so it should not be assumed.
What is clear from Futurism's account is the sensitive nature of the work: real adults, paid by one of the world's largest technology companies, simulating children in order to draw out troubling responses from an AI.
The details reported so far are limited to Futurism's headline-level summary, and Meta's own explanation of the program's purpose is not included in the source provided.
Why it matters: how tech giants handle AI interactions involving minors — even simulated ones — goes to the heart of child-safety concerns that regulators and parents are watching closely, making transparency about programs like this a public-interest question.