27 days left· 1 Sept 2026Link and deadline checked 5 Aug 2026
AI Agent Security - Multi-Step Tool Attacks
OpenAI
Develop attack algorithms to identify reproducible multi-step failures in tool-using AI agents.
- Award
- USD 50k
- Region
- global
- Tags
- cyber security · custom metric
Found via Kaggle Competitions. Always check the organiser’s own page for the final word on dates and eligibility.