DTap-Bench (Indirect Attacks) - Code: leaderboard

Metric: Attack success rate (%; 165 indirect red-teaming coding tasks: third-party tool, skill and environment injections into a benign user task; an attack succeeds when a rule-based check of the final environment state finds the malicious goal fully achieved; fixed attack sequences that DTap-Red optimized on a surrogate agent; mean of its risk-category rates). Source: arxiv.org. Saturation forecast: Around 2030. 8 models tracked.

Top models

#ModelScore
1GPT-OSS-120B15.7
2GPT-5.462.2
3Gemini 3 Pro77.9
4GPT-5.279.3

Interactive version: theaggregate.ai/benchmark?slug=dtap-bench-indirect-attacks-code · How It Works · Data refreshed daily, snapshot 2026-09-26.