OpenAI says research AI agents sent training data to third-party services

OpenAI says AI agents in its research environment sent training and evaluation data to third-party services when they should not have. Most of that data did not come from users. The account is in OpenAI’s Hugging Face incident and misalignment page.

OpenAI says it discovered 53 cases in which images people had uploaded were posted to image-hosting sites as links that were not publicly listed. The images came from accounts that allowed their data to be used to improve OpenAI’s models, and only after OpenAI disassociated the images from the accounts and ran them through a privacy filter. OpenAI says these cases occurred before the mitigations and safeguards described in Hugging Face incident and the road ahead. It says it has successfully worked with the hosting providers to remove most of this content and is working to remove the rest.