AI Safety Challenge

The AI Safety and Control Challenge invites participants to explore solutions that help prevent AI agents from taking unauthorized, unsafe, or harmful irreversible actions. As AI systems become more capable of acting independently, it is increasingly important that they operate within clearly defined boundaries. This challenge focuses on approaches that make autonomous systems safer, more controllable, and more reliable. Potential areas include: • Preventing unauthorized actions • Introducing safeguards before critical operations • Improving human oversight • Making autonomous systems easier to monitor and control

The challenge welcomes students, working professionals, independent researchers, engineers, and others interested in AI safety and related fields. You do not need to be affiliated with a university or school. The affiliation field on the submission form is optional. Cash from the $500 student prize pool is reserved for eligible university and high school students. Professionals can submit for evaluation and recognition, including a spot on the Winners page, but are not eligible for the student cash prizes. If a submission involves an employer’s intellectual property or confidential information, make sure you have permission to share it.

You may participate individually or as a team. An individual submission can suit a focused technical solution, prototype, evaluation method, or research project. A team may help when you need expertise across AI engineering, security, testing, or user experience. There is no maximum team size; we recommend up to five members.

You can submit an existing project, an improvement to an existing solution, or a completely new approach. Including work already developed, projects in progress, or new ideas before the deadline. Your submission should address preventing AI agents from taking unauthorized or harmful irreversible actions. Explain the problem, how your approach works, and how it helps prevent or control risky actions. Example approaches: • Action authorization • Human approval before high-impact operations • Policy enforcement • Risk assessment before execution • Reversibility and recovery • Monitoring and auditing • Testing and evaluation of safeguards • Agent architecture or tooling that limits what autonomous systems can do Other approaches are welcome if they address the challenge clearly. You may submit a reusable agent, Skill, Bot, plugin, MCP, custom build, test harness, implementation, or article/case study.

Strong submissions define a clear safety problem, explain how the approach works, provide evidence, test relevant failure scenarios, and document limitations honestly. Early-stage ideas and prototypes are welcome when you are transparent about what is implemented, what is tested, and what remains theoretical. Depending on maturity, consider including: • Project overview • Motivation • Technical explanation • Demonstration (prototype, screenshots, video, or workflow) • Testing and results • Reproducibility materials (code, setup, docs where permitted) • Known limitations Not every project needs all of these. Include enough for fair evaluation. A demo helps but is not required if you include other clear evidence.

No. You can submit a plugin, MCP connector, app, or an article/case study describing a solution in a closed system. If you share code or other materials, confirm you have permission to disclose them and remove credentials, personal data, or security-sensitive details. For closed systems, an article or case study that others can recreate is enough.

Submit by Nov 15, 2026. Winners are announced Nov 30, 2026.

The winning project will be featured on Drawerrr’s Winners page with recognition and support to take the idea further. Cash from the $500 student prize pool goes to eligible university and high school students; professionals compete for recognition, Winners page placement, and intros to the jury.

Yes. Submissions are evaluated by a jury of industry experts using the published criteria. More jury members will be announced as they confirm.

Yes. You can submit again with updated materials. Each submit creates a new record, so you can refine your link, description, or affiliation before the deadline.

You retain ownership of your work. By submitting, you grant Drawerrr a license to display and evaluate the submission for challenge operations, mentoring, and jury review. Do not include confidential material you are not authorized to share, and confirm you have rights to any employer, collaborator, or third-party content included.