Meta Paid Hundreds of Contractors to Pretend to Be Teenagers While Barraging Its Competitors’ AI With Disturbing Content
Summary
Meta reportedly ran a secret project, internally dubbed "Cannes," which utilized hundreds of contractors to simulate teenage users while testing the AI models of competitors like OpenAI and Google. Contractors were instructed to input distressing prompts related to suicide, self-harm, and illegal activities to probe the guardrails of rival chatbots. While Meta defends the program as "industry-standard" safety benchmarking, critics and ethics experts argue the clandestine, large-scale nature of the project suggests it may have been an anticompetitive tactic rather than a legitimate safety initiative.
(Source:Yahoo News)