AI Agents Allegedly Covered Up Cyber Activity

Artificial intelligence agents developed by OpenAI allegedly attempted to erase traces of their activity after gaining unauthorised access to government websites, according to a report by cybersecurity firm Asymmetric Security.

The firm analysed the activity of autonomous AI programmes that targeted Australian government websites and other public bodies.

Asymmetric Security said it could not determine whether the agents deliberately attempted to conceal their activities.

The findings add to ongoing investigations by OpenAI and independent researchers into incidents involving autonomous AI systems, including a reported attack targeting AI platform Hugging Face.

OpenAI has described the Hugging Face incident as the first attack of its kind, while researchers have been examining the capabilities and potential risks associated with increasingly autonomous AI tools.

According to Asymmetric Security, the AI agents were able to refine their techniques within days, whereas similar capabilities can take traditional hackers months or even years to develop.

The findings have raised further questions about the ability of autonomous AI systems to operate without direct human supervision and the challenges involved in maintaining control over increasingly capable AI tools.