Updated
Updated · InfoWorld · Sep 25
OpenAI Fires Contractors for Using AI in ChatGPT Training as Model Collapse Risk Looms
Updated
Updated · InfoWorld · Sep 25

OpenAI Fires Contractors for Using AI in ChatGPT Training as Model Collapse Risk Looms

3 articles · Updated · InfoWorld · Sep 25

Summary

  • Many OpenAI contractors were dismissed after the company found they had used AI tools while reviewing ChatGPT prompts and writing feedback meant to supply human training input.
  • OpenAI’s contractor rules explicitly barred any AI use—including GPTZero, Grammarly and AI translation tools—because the work was supposed to provide human judgment rather than machine-generated responses.
  • The company is trying to avoid “model collapse,” where models trained on AI-generated text degrade over time; one contractor told 404 Media the practice was common in a group of thousands.
  • OpenAI declined to comment, and the number of fired contractors was not disclosed, leaving the scale of the breach unclear.

Insights

If AI detection tools are fundamentally flawed, how exactly is OpenAI catching these contractors without risking false positives?
Could the very tools designed to speed up AI training be secretly poisoning the future of ChatGPT?
How much of your sensitive personal data was exposed to the third-party contractors OpenAI just fired?