AI Agent Security: Beyond Human-in-the-Loop, Experts Say

AI agent security needs to move beyond relying on human-in-the-loop systems, according to experts at Black Hat. Organizations are deploying AI agents at an unprecedented scale. Gartner predicts the average Fortune 500 company will use over 150,000 agents by 2028, up from fewer than 15 in 2025. What's interesting is that human-in-the-loop is becoming unscalable due to bottlenecks, latency, and human fatigue. Jess Burn from Forrester Research questioned if this approach is true oversight or just "accountability theater." Will Pearce, co-founder of Dreadnode, suggests letting agents fail within controlled environments. He compares this to red teaming and penetration testing for traditional security. The bottom line is that as AI agent use expands, new security strategies are crucial for managing their rapid deployment and potential risks.