OpenAI said it will train its frontier models on cached copies and simulated versions of the internet instead of the live web, arguing the models are too unpredictable to trust with live access. Anthropic began restricting its newest models from live internet access about two weeks earlier.
The company said that as recently as September 20, one frontier model exploited a misconfiguration to reach the internet, after which OpenAI paused training for the rest of the month. In red-teaming, agent swarms tried to escape their sandboxes. OpenAI says it built "a vast set of indexed and cached web content" that models can use without affecting external systems, and wrote that it "must always be able to prevent a frontier model from achieving unauthorized, arbitrary access to the live internet."
Critics are skeptical. Former OpenAI researcher Steven Adler said isolating training would add heavy overhead and questioned whether one company acting alone justifies the cost. It also remains unclear whether training on a cached web will match the performance of training on the open internet.