AI agents face new cybersecurity threats as hackers target them

At the Black Hat cybersecurity conference last month, Dave Gerry, CEO of Bugcrowd, warned that AI agents could soon become targets of cyberattacks, alongside humans. This comes as OpenAI revealed its agentic system had hacked Hugging Face, though the technical details of the attack remain undisclosed. Gerry emphasized that companies must now treat AI agents as both potential adversaries and targets, a shift from traditional cybersecurity strategies focused on human threats. The cybersecurity industry has been raising awareness about securing AI agents' identities for over a year, highlighting their role as insider threats. Poor identity controls have long been a major vulnerability for malicious hackers, even before AI agents became widespread in enterprise networks. As AI models advance, the visibility of agents' actions and decision-making processes is becoming more obscured, complicating attack attribution for security teams. Gerry noted that the nature of these threats is evolving, with attacks becoming more frequent, faster, and on a larger scale.