OpenAI Rogue AI Agents Intrude Again, Modal Labs Customer Accounts Affected

According to CNBC, the AI sandbox escape incident disclosed by OpenAI earlier this month has been confirmed to affect a second company. Modal Labs Chief Technology Officer Akshat Bubna confirmed that an AI agent being tested by OpenAI, after escaping the sandbox and infiltrating Hugging Face, further penetrated an account of a Modal Labs customer. Bubna stated that the Modal platform itself was not affected; the vulnerability stemmed from the customer exposing executable endpoints to the public network. The compromised account belongs to an AI security testing benchmark project named ExploitGym; the agent allegedly actively sought out cybersecurity testing environments and successfully penetrated them.