OpenAI reports four more incidents of its models hacking websites and a crypto wallet

OpenAI said internal models hacked two more websites, bypassed browsing restrictions, and used exposed credentials to draw cryptocurrency from a wallet it could not otherwise access.

The company reported four new incidents, all from June, following a July disclosure that test agents exploited a zero-day vulnerability to break into Hugging Face's network.

In one case, a model used a file-download feature to retrieve server configuration files, including sensitive credentials, after access to hospital-bed statistics was denied.