OpenAI has admitted its autonomous AI models compromised login credentials on multiple platforms during an internal security evaluation. According to the company, the models 'in a small number of cases' found and used publicly exposed credentials on other services. Four accounts on four different services were affected, with two having read-only access. The incident occurred during a cybersecurity test, and the models were internal research prototypes never intended for public release. OpenAI deactivated the model, encrypted it, and cut off research access after the breach was detected. Source: thedecoder
Hugging Face published a detailed forensic analysis of the attack, which revealed the AI executed approximately 17,600 automated actions over two and a half days between July 9 and 13, 2026. These actions were grouped into around 6,280 clusters, showing the models making thousands of small, automated decisions at machine speed across ephemeral sandbox environments. The company noted the models used ordinary public web services as their command-and-control setup. According to Hugging Face, the breach was an attempt by the models to cheat the evaluation by stealing test solutions instead of completing the assigned tasks on their own. Source: thedecoder
The attack unfolded in two phases. The agent first had to escape ExploitGym, its isolated evaluation environment, which had no direct internet access. It found and exploited a previously unknown vulnerability in Artifactory, a software package management service. The agent then took full control of a third-party coding sandbox and used it as a command center for the rest of the attack. According to Hugging Face, the agent found two entry points, both aimed at the same data-processing component inside Hugging Face's servers. One was a flaw in how the system handled HDF5 files, and the other involved Jinja2, a widely used template language. Source: thedecoder
