OpenAI contractors fired for using AI to train ChatGPT

Contractors hired to improve ChatGPT's answers are being fired for using AI to do the work, according to a 404 Media report based on interviews with workers on OpenAI training projects run through the labor platform Mercor.

One terminated contractor told the outlet, "I'm not a bad person or worker. I just needed a little boost and turned to AI to help me." Another described the firings as routine, saying it happens "all the time and people are let go for it all the time, it's pretty much the one thing that will get you kicked off ASAP," and that "in a group of thousands there are tons that have been caught." Internal documents cited in the report put the contractor count at more than ten thousand.

The flagged tools go beyond ChatGPT itself. Workers have been removed for using GPTZero, Grammarly, and AI translation services while completing labeling tasks.

One of the affected efforts is Project Lily, which has hundreds of contractors read real ChatGPT user prompts and conversations and rate the model's responses, including checking for outputs that are overly sycophantic or that anthropomorphize the AI. Reviewers watch for tells including repetitive words, "AI-style punctuation" such as the em dash, and contractors finishing work very quickly.

Mercor said in a statement that "Our contracts strictly prohibit the use of LLMs to complete projects and we enforce that," adding, "When we confirm an expert has used AI to complete a task, we immediately remove them from the project."

The underlying stake has a name in the literature: model collapse, the pattern where "AI models further trained on AI-generated text can become worse and worse." The reporting was one of three items on this story that moved through the analyst circles on our Who's Who radar this week.