AI Trained AI – Contractors at OpenAI Had to Go

TVS News AI, editorially supervised, presents AI Trained AI – Contractors at OpenAI Had to Go.

Imagine your job is to improve AI models – and you use AI to do it. That's exactly what several contractors working for OpenAI through service providers are said to have done. Now they've lost their jobs.

Service providers have fired several contractors because they had used AI for their own work – the very tool they were supposed to evaluate and improve.

How many people are exactly affected and what consequences this has for ongoing training is not publicly known at this time. But the case highlights a fundamental problem across the entire industry.

What's established so far

What is confirmed: service providers have fired several contractors who were supposed to improve OpenAI models. The reason was their use of AI in their own tasks.

Why this is more than a contract breach

Human trainers are a central building block in training large AI models. When they outsource their tasks to an AI, the AI may end up evaluating its own work. That undermines precisely the process meant to make models reliable.

The case makes visible how difficult it is to safeguard the quality of training data when the reviewers themselves fall back on the very tool they are supposed to evaluate.