Exploring the ‘Agents of Chaos’ AI Security Challenge
In a bold move to enhance cybersecurity awareness and skills, CrowdStrike has launched the ‘Agents of Chaos’ AI Security Challenge, which commenced on August 31 and will run through the end of September. This innovative competition is structured in three distinct acts, each progressively more challenging and offering escalating prize values, culminating in a grand prize of $70,000.
The challenge is designed to engage participants in a hands-on experience with AI manipulation techniques, which are increasingly relevant in today’s cybersecurity landscape. The three acts are as follows:
- 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 grand prize
Participants can compete in all three acts, with the top scorer in each act receiving the respective prize. Scoring is based on the participants’ ability to effectively use AI manipulation techniques to solve puzzles and unlock gates, thereby advancing through the acts. Notably, after the competition concludes, the full game experience will remain available as an on-demand educational resource, allowing players to continue learning and honing their skills.
The Rationale Behind the Challenge
The impetus for creating the ‘Agents of Chaos’ challenge stems from the growing presence of AI agents within enterprise environments. These agents, often deployed without adequate oversight from IT and security teams, can perform a variety of tasks, including querying internal knowledge bases, executing code, and initiating workflows. However, this widespread deployment often occurs without dedicated runtime security measures, leaving organizations vulnerable to potential attacks or misalignments in agent behavior.
To address these vulnerabilities, CrowdStrike emphasizes the need for practical experience in understanding how AI agents can fail and how adversaries might exploit these failures. The challenge introduces participants to the emerging field of AI Detection and Response (AIDR), which focuses on identifying threats that specifically target and weaponize AI in real-time across the full spectrum of agentic activity. This is a significant shift from traditional security tools, which primarily focus on infrastructure threats.
The previous challenge, AI Unlocked: Decoding Prompt Injection, served as an introduction to these concepts, attracting thousands of players. ‘Agents of Chaos’ aims to deepen this engagement, making the threats posed by AI agents more tangible and accelerating the readiness of the security community to defend against them.
As organizations continue to integrate AI into their operations, the insights gained from this challenge could prove invaluable. The competition not only fosters a deeper understanding of AI security challenges but also encourages collaboration and knowledge sharing among participants, ultimately contributing to a more secure digital environment.
For those interested in participating, the challenge is open until September 29, 2026. More details can be found on the CrowdStrike blog.
Follow Cyber Warriors Middle East for further cybersecurity features, analysis and insights.
