{"id":132113,"date":"2026-08-06T21:10:20","date_gmt":"2026-08-06T21:10:20","guid":{"rendered":"https:\/\/www.europesays.com\/ai\/132113\/"},"modified":"2026-08-06T21:10:20","modified_gmt":"2026-08-06T21:10:20","slug":"crowdstrikes-100k-agents-of-chaos-contest-turns-ai-red-teaming-into-a-game","status":"publish","type":"post","link":"https:\/\/www.europesays.com\/ai\/132113\/","title":{"rendered":"CrowdStrike\u2019s $100K Agents of Chaos Contest Turns AI Red Teaming Into a Game"},"content":{"rendered":"<p>\t\t\tTL;DR \u2014 Key Takeaways<br \/>\nCrowdStrike is offering $100,000 in total prizes through an international competition focused on AI agent security.<br \/>\nPlayers must use prompt injection and other red-teaming techniques to manipulate live AI agents.<br \/>\nThe competition rewards both successful attacks and prompt efficiency, with points deducted for excessive token use.<\/p>\n<p>CrowdStrike has announced a $100,000 international competition that challenges participants to manipulate live AI agents using prompt injection and other red-teaming techniques.<\/p>\n<p>The competition takes the form of a virtual game called AI Unlocked: Agents of Chaos, created in collaboration with Amazon Web Services. Players take on a fictional mission to stop a shadow group from deploying malicious AI. To do that, they must manipulate the group\u2019s own agents into revealing information and taking actions they were not authorized to perform. Their scores also depend on efficiency, with points deducted for every token used in their prompts.<\/p>\n<p>The contest opens Aug. 31 and is divided into three acts. The first runs through Sept. 7 with a $10,000 prize, the second ends Sept. 14 and awards $20,000, and the final act runs from Sept. 15 through Sept. 29 with a $70,000 prize. The game is available online internationally, with an in-person option to play at CrowdStrike\u2019s Fal.Con conference. <a href=\"https:\/\/www.crowdstrike.com\/en-us\/platform\/falcon-aidr-ai-detection-and-response\/agents-of-chaos\/\" rel=\"nofollow noopener\" target=\"_blank\">Pre-registration<\/a> is open, and contestants can replay completed puzzles to refine their prompts. CrowdStrike said the highest score recorded when each act closes will receive that act\u2019s prize.<\/p>\n<p>CrowdStrike\u2019s challenge turns a real security problem into a game. AI agents are becoming more autonomous and can run commands, use tools and act with valid credentials, making it harder to tell whether unusual behavior comes from an attack or from the agent doing what it was told.<\/p>\n<p>That problem is already showing up in CrowdStrike\u2019s threat data. Its <a href=\"https:\/\/securityboulevard.com\/2026\/08\/report-ai-has-become-both-weapon-and-target-in-modern-cyberattacks\/\" rel=\"nofollow noopener\" target=\"_blank\">2026 Threat Hunting Report<\/a> found that activity initiated by AI agents generated detection leads at 2.5 times the rate of human activity. The figure reflects leads identified through CrowdStrike\u2019s own telemetry between July 2025 and June 2026, not confirmed attacks, but it shows the growing volume of agent activity security teams now have to sort through.<\/p>\n<p>During a media roundtable about the report, Adam Meyers, CrowdStrike\u2019s senior VP of counter adversary operations, explained how the company identifies agent activity by tracing it back through the process tree to tools such as Codex or Claude. Once an agent starts a task, CrowdStrike can trace commands, dependency downloads and other activity back to the agent that initiated them.<\/p>\n<p>Identifying the activity is one thing, but understanding what it means is harder. Meyers said CrowdStrike has seen legitimate AI tools take a different action when a security control blocks their initial attempt, not necessarily to evade detection, but because the agent treats the control as an obstacle to completing its assigned task.<\/p>\n<p>\u201cI\u2019m trying to do this, but the security tool got in my way, so I\u2019ll just go around it, right? The agent doesn\u2019t know that that\u2019s bad, or that it shouldn\u2019t do that,\u201d Meyers said. Each workaround can trigger additional detections, creating a cascading flow of events that security teams must sort through.<\/p>\n<p>The Agents of Chaos challenge will give defenders a controlled environment to practice turning an agent against its intended purpose. When that same behavior appears inside a live environment, authorized access can make manipulated activity seem routine. The game may be fictional, but the skills it requires are quickly becoming part of security teams\u2019 daily work.<\/p>\n<p>Frequently Asked QuestionsWhat is AI Unlocked: Agents of Chaos?<\/p>\n<p>It is a virtual cybersecurity game created by CrowdStrike in collaboration with Amazon Web Services. Participants manipulate live AI agents to stop a fictional group from deploying malicious AI.<\/p>\n<p>How much prize money is available?<\/p>\n<p>CrowdStrike is offering $100,000 across three acts: $10,000 for the first, $20,000 for the second and $70,000 for the final act.<\/p>\n<p>When does the competition take place?<\/p>\n<p>The competition begins Aug. 31. The first act ends Sept. 7, the second ends Sept. 14 and the final act runs from Sept. 15 through Sept. 29.<\/p>\n","protected":false},"excerpt":{"rendered":"TL;DR \u2014 Key Takeaways CrowdStrike is offering $100,000 in total prizes through an international competition focused on AI&hellip;\n","protected":false},"author":2,"featured_media":132114,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[6],"tags":[179,405,65842,7537,330,313,8922],"class_list":["post-132113","post","type-post","status-publish","format-standard","has-post-thumbnail","category-agentic-ai","tag-agentic-ai","tag-ai-agents","tag-ai-unlocked-agents-of-chaos","tag-artificial-intelligence-agents","tag-crowdstrike","tag-cybersecurity","tag-red-teaming"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/132113","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/comments?post=132113"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/132113\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media\/132114"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media?parent=132113"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/categories?post=132113"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/tags?post=132113"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}