OpenAI says it will train its frontier AI models using cached web content and simulated online environments rather than the live internet. The company says this approach will let models work with broad web information while keeping external systems separate from training.
To support the change, OpenAI has built an indexed collection of cached web content for models to use. The company also says its training environments must prevent models from gaining unauthorized access to the live internet.
The shift follows security testing that highlighted the need for stronger safeguards around model training. OpenAI says it paused training work in September to develop and test additional protections.
Using offline web environments could reshape how advanced AI systems are trained, balancing access to information with more controlled digital boundaries.
