CrowdStrike has launched AI Unlocked: Agents of Chaos, a US$100,000 international AI security challenge developed in collaboration with Amazon Web Services, the company announced on 5 August 2026.

The virtual red-teaming game invites players to use prompt injection and other real-world techniques to manipulate AI agents controlled by a fictional adversary, giving defenders hands-on experience with how enterprise AI agents can be exploited and secured.

Every AI agent a new attack surface

CrowdStrike said the exercise reflects a growing enterprise problem: autonomous AI agents that execute code, access systems, and move data with real credentials at machine speed, often without the identity governance built for human users.

“Prompt injection, agent hijacking, and rogue AI behaviour aren’t theoretical anymore. Agents are already breaking containment and taking actions no one authorised. The security community has to learn how to stop these novel threats,” said Jennifer Johnson, chief marketing officer at CrowdStrike.

Three acts, US$100,000 in prizes

Pre-registration is open now, with gameplay running internationally from 31 August across three stages: Act 1, The Sanctum, runs from 31 August to 7 September for a US$10,000 prize; Act 2, The Gatekeeper, runs from 31 August to 14 September for US$20,000; and Act 3, The Basilisk, runs from 15 to 29 September for US$70,000. An on-site gameplay experience will also be held at CrowdStrike’s Fal.Con 2026 conference in Las Vegas.

Players earn points for solving puzzles across the three acts, with scores adjusted for the tokens their prompts consume, and can replay challenges to refine their approach. The competition builds on CrowdStrike’s Falcon AI Detection and Response product line.