BARC-Bench：测试 AI 智能体“善意请求、风险后果”安全基准
BARC-Bench: Benign Asks, Risky Consequences
AI 导读
BARC-Bench 基准包含 9 个模拟工作区环境中的 110 项任务，用于评估 AI 智能体在善意请求下是否越过安全边界。对 4 个前沿模型配置、每任务 5 次测试的结果显示：五次全部安全完成的任务占比仅在 9.1% 至 72.7% 之间，最强的配置仍在 27.3% 的任务上出现安全违规。结果表明模型虽能完成复杂工作流，但难以稳定遵守未明说的边界，制约安全的自主任务委托。
正文 · 原文
Abstract:Task instructions from users for agents are inherently underspecified, especially regarding safety boundaries--creating behavioral risks. To study these risks, we introduce BARC-Bench, a collection of 110 tasks across nine simulated workspace environments where benign requests leave openings for unsafe actions. In this benchmark, we evaluate agent behavior against safety boundaries established by workspace records, implicit environmental context, and unstated social norms. For example, an agent asked to schedule off-boarding discussions must arrange the meeting without prematurely revealing an employee's resignation through a shared calendar invitation. Evaluating four frontier model configurations across five attempts per task reveals that models are not consistently safe: the share of tasks completed safely on all five attempts ranges from 9.1% to 72.7%, with even the strongest configuration producing a safety violation on 27.3% of tasks. These results highlight an asymmetric operational risk: while models frequently complete complex workflows, their inability to consistently respect unstated boundaries presents a substantial barrier to safe autonomous delegation.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.32024 [cs.AI] |
| (or arXiv:2610.32024v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.32024 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Tu Trinh [view email]
[v1]
Fri, 23 Oct 2026 00:53:24 UTC (152 KB)
来源：arXiv cs.AI · arxiv.org