Seven cybersecurity veterans and elite students clashed at the National Collegiate Cyber Defense Competition in Las Vegas, where AI agents—part of an Anthropic-backed team—faced off against human blue teams. The red team, led by Alex Levinson, aimed to breach San Antonio's network using custom malware, while AI bots like Claude Code automated tasks, sometimes misfiring (e.g., installing malware on their own machine). Dan Borges, a former Uber engineer, noted AI can speed up tasks but struggles with nuanced decision-making, as bots 'hallucinated' non-existent network activity.

The AI team, comprising 32 agents versus each blue team's eight students, initially ranked last due to a network outage but improved once online. Expert David Cowen highlighted that AI's real power lies in parallel processing (e.g., 'going fast, going wide'), yet it remains unreliable without human oversight—'the bots are only as good as the people using them.'.