OpenAI has launched a new opt-in security feature for ChatGPT accounts, Decrypt reports.
According to Decrypt, the feature hinges on passkeys. That means users must use passkeys rather than relying on more traditional sign-in or verification flows.
Decrypt also says OpenAI limits recovery options. The goal is to reduce the odds of account takeovers via weak recovery paths, but users should expect fewer ways to regain access if they lose their credentials.
Finally, Decrypt reports that the feature excludes chats from training. In other words, opting in changes how OpenAI may use those conversations compared with the baseline settings.
OpenAI’s move is a straightforward security tightening, not a guarantee of safety. As Decrypt frames it, the tradeoff is clear. Stronger sign-in and recovery constraints in exchange for privacy changes around training.