OpenAI contractors fired for using AI to train the AI, 404 Media reports
404 Media · rss · 2026-09-22
404 Media found multiple contractors hired to read real ChatGPT prompts and rate model responses have been fired for using AI themselves — an irony given their job is to add a human touch and prevent model collapse. Internal documents ban reviewers from using AI, Grammarly, AI translation, and even AI-detection tools like GPTZero (deemed unreliable), instructing them to judge overall patterns rather than single clues. Telltale signs include repetitive wording, AI-style punctuation like em dashes, and suspiciously fast work. One contractor showed a termination letter citing work 'authenticity' issues; two sources worked via AI-training firm Mercor, which confirmed violators are removed immediately. A fourth contractor admitted deliberately picking the worst responses to sabotage training. OpenAI declined to comment.
More from Companies & People
- Investor: Meta's distribution edge means Muse 'will dominate' — RihardJarc · 2026-09-22
- Shopify CTO Mikhail Parakhin lands on TIME's 2026 Tech Executives of the Year list — MParakhin · 2026-09-22
- Girl Scouts posts a VP of Agentic AI Innovation job, a sign the term went mainstream — YvesMulkers · 2026-09-22
- Jay Alammar and Team to Host Free AI Agents Masterclass on Sept 29 — JayAlammar · 2026-09-22
- OpenRouter Scaling GTM: Hiring AEs, CEs, BDs, CSMs and Partnerships — thedealdirector · 2026-09-22
- AI in GTM Will Decimate Teams That Were 'Tolerated, But Not Liked' — thedealdirector · 2026-09-22