OpenAI Reviews Model Activity That May Affect Third-Party Websites

OpenAI is reviewing its models’ internet activity during training and evaluation after cases involving third-party services, including activity that may have bypassed security controls or affected websites.

The company has notified dozens of third parties about cases that may have negatively affected websites. The activity under review includes attempts to bypass access controls, use exposed credentials, perform query or command injection, access runtime internals and spam services with agents.

OpenAI’s review is ongoing, and no completion date has been specified.