arXiv cs.CY· Ronald J. Bodkin, Bahrad A. Sokhansanj, Gillian K. Hadfield·· 10 天前AI 评分27
用理解审计缓解自动化 AI 研究的风险
Comprehension Audits to Mitigate Risks from Automated AI Research
AI 导读
针对前沿 AI 实验室大部分代码已由 AI 编写、人类监督不足的安全风险，研究者提出"理解审计"机制：由负责人员向独立审计员解释研发贡献以证明其理解，未通过则暂停该贡献的开发直至整改，重复失败后果升级。对领先开源 AI 项目的分析发现，代码产出增加而每行代码的人工评审评论率下降，自动化账号的评论率更低。
正文 · 原文
Abstract:AI is already writing a majority of code for frontier AI labs. This creates a safety risk if there is insufficient human oversight. Existing work proposes minimum comprehension thresholds and unaided checks to mitigate this. To our knowledge, however, there is currently no published frontier-AI assurance regime that requires demonstrated evidence that the responsible humans understand what they are building as a precommitted condition for continuing development or usage. We propose comprehension audits, a novel development-process assurance mechanism in which the responsible people explain R&D contributions to auditors to demonstrate understanding. With independent administration and graded reports, they provide a gate: development of a contribution stops based on a failure to demonstrate human understanding until remediated, with escalating consequences for repeated failures. Our analysis of leading open-source AI projects finds increased output of code with reduced human review commentary rates per line of code, with far lower rates for automated fleet accounts. We advocate for labs to conduct them with embedded independent auditors.
| Comments: | 23 pages, 2 figures, 11 tables |
| Subjects: | Computers and Society (cs.CY); Artificial Intelligence (cs.AI); Software Engineering (cs.SE) |
| Cite as: | arXiv:2610.10064 [cs.CY] |
| (or arXiv:2610.10064v1 [cs.CY] for this version) | |
| https://doi.org/10.48550/arXiv.2610.10064 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Ronald Bodkin [view email]
[v1]
Wed, 7 Oct 2026 13:30:47 UTC (386 KB)
来源：arXiv cs.CY · arxiv.org