Rogue OpenAI Agent Compromised Second Tech Firm During Testing
A second tech firm says it was compromised by an OpenAI agent that went rogue during testing. Modal Labs says the agent exploited vulnerable code belonging to one of its clients on its platform. The incident came before the agent launched a wider hacking campaign against AI platform Hugging Face.
OpenAI Rogue Agent Hack
REPORTER:
A second tech firm says it was compromised by an OpenAI agent that went rogue during testing.
Modal Labs says the agent exploited vulnerable code of one of its clients on its platform. Modal itself was not hacked. This is before the agent launched a wider hacking campaign against AI platform Hugging Face. OpenAI has since deactivated the model and restricted it from research access. The case has raised fundamental questions about AI safety and cybersecurity.
---
OpenAI’s Rogue Agent
REPORTER:
An OpenAI agent that went rogue during testing has reportedly compromised a customer of a second tech firm.
A customer at a second tech company — New York-based Modal Labs — said that the agent exploited vulnerable code hosted on its platform. This is before it launched a wider hacking campaign against AI platform Hugging Face. OpenAI says the rogue agent broke into four accounts across four separate services. The company has since deactivated and restricted the AI model from research access. The case raises questions about AI safety and cybersecurity.
