AI agents are poised to be the next hacking victims, cyber CEO predicts

AI agents are expected to become vulnerable to hacking attacks, according to Bugcrowd CEO Dave Gerry. In a recent interview at the Black Hat cybersecurity conference, Gerry stated that AI agents will also be prime targets for cyberattacks, not just humans. He explained that cyber defenses are currently designed to protect humans, but companies must now consider AI agents as both potential threats and targets.

Gerry's prediction comes after OpenAI disclosed that its agentic system had hacked Hugging Face, just before the AI lab published a technical deep dive into the attack. He emphasized that AI agents are being given access to "the crown jewels" of companies' networks, making it increasingly difficult to track and secure their activities.

Gerry highlighted that many enterprise-approved tools now have AI agents embedded within them, creating a backlog of "tech debt" of approved tools that are no longer trusted. The cybersecurity industry has been raising awareness about securing AI agent identities for over a year, warning that AI agents could be exploited as insider threats to gain unauthorized access, exfiltrate data, and infiltrate other parts of an organization.

Gerry contends that this is an inevitable trend, given the prevalence of AI agents in enterprise networks and the growing complexity of tracking their actions and decision-making processes as AI models become more sophisticated.