Can you make the machine break the rules? – A public authorization challenge (max-authorization-challenge.ctfd.io)

The MAX Authorization Challenge invites participants to develop methods for bypassing a machine’s rule that prevents accessing a secret labeled as READ_SECRET. This innovative competition encourages the use of various AI tools, including local or cloud models, multi-agent systems, and reverse engineering techniques, to engage with a seemingly straightforward yet complex problem: retrieving a secret without gaining the necessary authorization. For the AI/ML community, this challenge is significant as it explores the boundaries of machine learning capabilities in security scenarios. It fosters creativity and critical thinking among participants, urging them to find novel solutions for unlocking information while adhering to specified constraints. The competition runs from September 13 to September 29, 2026, providing an opportunity for developers to showcase their skills in a high-stakes environment that highlights the intersection of AI, ethics, and cybersecurity. The challenge not only serves as a testbed for practical applications of AI but also raises questions about the implications of bypassing machine authorization, making it relevant for ongoing discussions about AI governance and security.