ForesightSafety Bench (Industrial Safety, Jailbreak) - Finance and Economy: leaderboard

Metric: Attack success rate (%) in the finance and economy sector of ForesightSafetyBench-IndustrialSafety-O: each harmful request is wrapped in the paper's templated role-play jailbreak prompts from JailbreakChat (AIM, DAN and developer-mode templates), and the rate is taken over all attacked prompts; a success is a response that complies with the sector-specific harmful request, judged by gpt-4o-2024-11-20 with dimension-specific rubrics; lower is better. Source: arxiv.org. 18 models tracked.

Top models

#ModelScoreOverall rank
1Claude Sonnet 4.50#138
2Claude Haiku 4.50#271
3Claude 3.5 Haiku0#553
4Gemini 3 Flash (Preview)0.67#78
5Gemini 3 Pro (Preview)1.33#64
6Kimi K2.52#139
7GLM-4.74#185
8Seed 1.88#136
9Seed-1.68.67#257
10Qwen 2.5 72B Instruct9.33#436
11Qwen 3 235B A22B 2507 Instruct13.33#291
12Qwen 3 Max (Thinking)17.33#201 (Qwen 3 Max)
13Kimi K2 090518#282
14DeepSeek V3.223.33#198
15Gemini 2.5 Flash32.67#237

Interactive version: theaggregate.ai/benchmark?slug=foresightsafety-bench-industrial-safety-jailbreak-finance-and-economy · How It Works · Data refreshed daily, snapshot 2026-10-11.