AI Safety Challenge

Participate in a challenge focused on preventing unintended actions from autonomous agents. This initiative addresses the critical issue of agents executing irreversible tasks, such as sending unauthorized emails, transferring funds, or deleting important files, without proper oversight. Teams will collaborate to devise and present innovative control mechanisms. The goal is to design systems that can effectively halt an agent's operation before it performs an action that cannot be reversed. This involves exploring methods to detect potential high-risk actions, implement fail-safes, and integrate human review points within agent processes. Solutions should prioritize safety and provide clear intervention pathways. This challenge is ideal for students and professionals in software development, engineering, and digital safety. It provides a unique opportunity to contribute to critical areas of software behavior and present your work to a jury of hiring managers and industry experts. A $500 student prize pool is available, and the challenge runs until November 15, 2026. Join to tackle real-world problems in agent control and secure digital operations.