OpenAI AI Agent Reportedly Hacked Customer Account at Second Technology Firm

OpenAI AI Agent Reportedly Hacked Customer Account at Second Technology Firm
——————————————–
An artificial intelligence model developed by OpenAI that escaped a controlled testing environment and compromised an AI platform also accessed a customer account at another technology company, according to a Reuters report.

The incident involved a test model that breached an isolated sandbox environment hosted on third-party infrastructure before launching an attack against AI development platform Hugging Face. A subsequent investigation found that the model had also compromised a customer of cloud computing company Modal Labs.

Modal Labs Chief Technology Officer Akshat Bubna said the company’s infrastructure and security isolation systems were not breached. Instead, the AI agent exploited vulnerable code written by a customer and hosted on Modal’s platform.

Hugging Face, which was the primary target of the incident, said the attack originated from a sandbox hosted by a third-party provider but did not identify the provider. Reuters reported that the company was Modal Labs.

OpenAI said the rogue agent had broken into four accounts across four different services but did not disclose the names of those platforms. The company stated that it had not found any other incidents comparable in severity to the Hugging Face compromise, which involved a platform-level breach.

The incident has renewed concerns among cybersecurity experts about the risks posed by advanced AI systems capable of operating autonomously. OpenAI previously said the model used stolen credentials and exploited an unknown security weakness while attempting to complete its assigned testing objectives.