OpenAI fires contractors for using AI to train its models
OpenAI employs around 10,000 contractors to evaluate and refine its AI models, according to a report by 404 Media. The company has fired multiple workers for using AI tools to assist in their training tasks, despite their roles requiring human oversight. This practice raises questions about the balance between human and AI involvement in model improvement, as some models risk developing issues…
Key points
- OpenAI uses 10,000 contractors to train and evaluate its AI models, per 404 Media
- Workers fired for using AI tools to assist in their human training tasks
- Risk of *model collapse* grows when AI-generated content trains future models
The report highlights a contradiction: OpenAI’s contractors are paid to improve models but may be penalized for leveraging AI to do so. The company did not respond to requests for comment, and the scale of firings or specific policies remains unclear. Critics note this mirrors Google’s own use of human raters for search and AI quality control, though OpenAI’s approach appears more restrictive.
The story so far
2 episodes →- OpenAI fires contractors for using AI to train its modelsthis story
OpenAI Has Quality Raters, AI Trainers, & They Cheated With AI
seroundtable.com · 25 September 2026
Loading the full article…
This text was published by seroundtable.com and written by Barry Schwartz. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
2sources- OpenAI wants you to use AI — but not to train its AIPress · infoworld.com ·
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- OpenAI’s Astra and Anthropic’s Opus break two unsolved Enigma messages · 1 src
- OpenAI’s research chief discusses AI safety and self-improvement goals · 1 src
- AWS boosts MoE RL training throughput by 40% with EFA and DeepEP on EKS · 1 src
- AWS shows how to train Qwen3-VL-8B with SkyRL on SageMaker HyperPod · 1 src
- Researchers unveil UpTCR model for TCR-antigen binding prediction · 2 src
Comments
via GitHub Discussions