Meta paid contractors to pose as children, flood rival chatbots with prompts about suicide, eating disorders, and abuse, then log every response in spreadsheets. The company calls it "industry-standard safety benchmarking," but the operation had no consent, no disclosure, and no shared findings: the four things that make red teaming legitimate. Host Emily Laird walks through the Wired investigation, the 45,000-prompt testing rounds, and the court testimony showing what Meta knew about its own failure rates while it was busy documenting everyone else's. This is the difference between a shield and a sword, and the paperwork says sword.────────────
Podden och tillhörande omslagsbild på den här sidan tillhör
Emily Laird. Innehållet i podden är skapat av Emily Laird och inte av,
eller tillsammans med, Poddtoppen.
