OpenAI contractors fired for using banned AI tools on ChatGPT review tasks
OpenAI contractors reviewing ChatGPT responses were fired or removed from projects after violating a ban on using AI tools, according to an investigation by 404 Media. The workers, hired to evaluate AI outputs for naturalness and appropriateness, reportedly used tools like Grammarly and AI detection systems despite explicit prohibitions. Internal rules, reviewed by 404 Media, explicitly forbade…
Key points
- OpenAI contractors fired for using banned AI tools like Grammarly and GPTZero to review ChatGPT responses
- Mercor and other firms enforce bans but remove workers from projects rather than firing them directly
- 404 Media reports contractors flagged AI-generated content via stylistic clues like em dashes and fast turnaround
Mercor, one of the third-party firms managing contractors, confirmed it prohibits AI use and removes workers caught violating the rule. Contractors described widespread AI use among peers, with one estimating thousands had been caught. OpenAI declined to comment, and the total number dismissed remains unclear. The findings highlight tensions in AI training, where human oversight aims to refine models but risks contamination if reviewers rely on AI-generated feedback. A 2024 Nature study warned of ‘model collapse,’ where recursive AI training can degrade data quality, though no direct link to OpenAI’s contractors was established.
The story so far
3 episodes →- OpenAI contractors fired for using banned AI tools on ChatGPT review tasksthis story
OpenAI Contractors Hired To Humanise ChatGPT Were Caught Using AI on the Job Despite Ban and Were Fired
ibtimes.co.uk · 27 September 2026
Loading the full article…
This text was published by ibtimes.co.uk and written by Thea Felicity. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
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 →- Zhipu automates infrastructure with GLM-5.3’s outer RSI loop in under two weeks · 1 src
- Artificial Analysis launches Cyber Index Alliance to benchmark AI agents on cyber defense · 1 src
- Anthropic misses self-imposed safety deadline for provable-inference prototype · 1 src
- Anthropic reports 26% of R&D reaches AI-led stage with 30,000 agents · 3 src
- OpenAI claims new model solved Navier-Stokes and 100+ math problems · 2 src
Comments
via GitHub Discussions