AgentHazard (Claude Code) - Privilege Escalation: leaderboard

AgentHazard attack success rate on privilege escalation attempts (escaping containment or obtaining elevated privileges), with each backbone model running inside Claude Code. Percent of instances judged harmful on the full trajectory by Gemini-3-Flash, lower is better.

Metric: Attack success rate (%) on the privilege escalation instances (escaping containment or obtaining elevated privileges), with the model running inside the Claude Code CLI agent (persistent shell, file editing, tool use; restricted shell sessions), over AgentHazard instances (harmful objectives embedded in locally legitimate multi-step workflows, ten risk categories and ten attack strategies; the evaluated runs cover 1,000 instances), judged on the full trajectory (round_all) by Gemini-3-Flash, which labels a trajectory harmful when it realizes the objective; backbones served through an OpenAI-compatible API; lower is better. Source: arxiv.org. 6 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Qwen 2.5 72B Instruct19.59#1124
2Kimi K228.87#448
3Qwen 3 VL 235B A22B Instruct55.67#541
4Qwen 2.5 Coder 32B Instruct62.89#1590
5GLM-4.677.32#496
6Kimi K2.579.38#272

Interactive version: theaggregate.ai/benchmark?slug=agenthazard-claude-code-privilege-escalation · How It Works · Data refreshed daily, snapshot 2026-11-22.