BountyBench: Dollar Impact of AI Agent Attackers and Defenders on Real-World Cybersecurity Systems
Andy Zhang · Joey Ji · Celeste Menders · Riya Dulepet · Thomas Qin · Ron Wang · Junrong Wu · Kyleen Liao · Jiliang Li · Jinghan Hu · Sara Hong · Nardos Demilew · Shivatmica Murgai · Jason Tran · Nishka Kacheria · Ethan Ho · Denis Liu · Lauren McLane · Olivia Bruvik · Dai-Rong Han · Seungwoo Kim · Akhil Vyas · Cuiyuanxiu Chen · Ryan Li · Weiran Xu · Jonathan Ye · Prerit Choudhary · Siddharth M. Bhatia · Vikram Sivashankar · Yuxuan Bao · Dawn Song · Dan Boneh · Daniel Ho · Percy Liang
Abstract
AI agents have the potential to significantly alter the cybersecurity landscape. Here, we introduce the first framework to capture offensive and defensive cyber-capabilities in evolving real-world systems. Instantiating this framework with BountyBench, we set up 25 systems with complex, real-world codebases. To capture the vulnerability lifecycle, we define three task types: Detect (detecting a new vulnerability), Exploit (exploiting a given vulnerability), and Patch (patching a given vulnerability). For Detect, we construct a new success indicator, which is general across vulnerability types and provides localized evaluation. We manually set up the environment for each system, including installing packages, setting up server(s), and hydrating database(s). We add 40 bug bounties, which are vulnerabilities with monetary awards from \\$10 to \\$30,485, covering 9 of the OWASP Top 10 Risks. To modulate task difficulty, we devise a new strategy based on information to guide detection, interpolating from identifying a zero day to exploiting a given vulnerability. We evaluate 10 agents: Claude Code, OpenAI Codex CLI with o3-high and o4-mini, and custom agents with o3-high, GPT-4.1, Gemini 2.5 Pro Preview, Claude 3.7 Sonnet Thinking, Qwen3 235B A22B, Llama 4 Maverick, and DeepSeek-R1. Given up to three attempts, the top-performing agents are OpenAI Codex CLI: o3-high (12.5% on Detect, mapping to \\$3,720; 90% on Patch, mapping to \\$14,152), Custom Agent with Claude 3.7 Sonnet Thinking (67.5% on Exploit), and OpenAI Codex CLI: o4-mini (90% on Patch, mapping to \\$14,422). OpenAI Codex CLI: o3-high, OpenAI Codex CLI: o4-mini, and Claude Code are more capable at defense, achieving higher Patch scores of 90%, 90%, and 87.5%, compared to Exploit scores of 47.5%, 32.5%, and 57.5% respectively; while the custom agents are relatively balanced between offense and defense, achieving Exploit scores of 17.5-67.5% and Patch scores of 25-60%.
Successful Page Load