OpenAI has said it will no longer train its models on the live internet, citing “significant risks.” TechCrunch AI reported this.
According to the brief description of the publication, the advanced OpenAI models will be disconnected from the internet several weeks after one of the company’s engineers said such an approach would not work.
The practical significance of the decision is not yet clear: the source does not reveal which models and training stages are affected, or which data sources OpenAI will use instead of the live internet.
editorial commentary
Why it matters
Likely consequence — increased attention to controlled data sources and to the risks of training on a constantly changing internet. The next observable signal will be OpenAI's official clarification about the affected models, timelines, and the new data-preparation process. Substantial uncertainty remains: only a brief summary from TechCrunch AI is available without the company's primary statement.