CrowdStrike has announced AI Unlocked: Agents of Chaos, a global AI security competition developed in collaboration with AWS that aims to help cybersecurity professionals understand and defend against emerging threats targeting AI agents.
The virtual challenge focuses on real-world attack techniques such as prompt injection, agent hijacking, and unauthorized AI behavior. Participants will attempt to manipulate AI agents controlled by a fictional adversary, gaining hands-on experience with vulnerabilities that can affect autonomous systems deployed in enterprise environments.
As organizations increasingly adopt AI-powered agents, CrowdStrike warns that these systems introduce new security risks. AI agents operate as machine identities capable of accessing enterprise resources, executing code, and handling sensitive data, creating new attack surfaces that traditional security tools were not designed to protect.
The competition consists of three acts, each requiring players to use red-teaming techniques to exploit AI agents, uncover information, and prevent a simulated malicious AI attack. Players earn points based on successful actions while minimizing token consumption, with top performers receiving cash prizes.
The challenge will be open worldwide starting August 31, with online participation available globally and an on-site experience at Fal.Con 2026. The prize pool totals $100,000, distributed across three stages:
- Act 1: The Sanctum (August 31–September 7) – $10,000 prize
- Act 2: The Gatekeeper (August 31–September 14) – $20,000 prize
- Act 3: The Basilisk (September 15–29) – $70,000 prize
CrowdStrike said the initiative is designed to help security teams better understand how AI agents can be manipulated at runtime and how organizations can strengthen defenses in the emerging era of agentic AI.
