AI Struggles in Cybersecurity: SecRespond Benchmark Reveals Limits

A new benchmark, SecRespond, reveals that AI agents consistently fail at detecting silent intrusions and crafting complete remediation plans during post-compromise investigations. While they excel in alert-driven responses, they falter when proactive hunting is required. The study highlights a critical gap in AI capabilities, emphasizing the need for human analysts in complex cybersecurity scenarios.