LectureRechercheRadarCadre d'investissement
Connexion / Inscription
← Retour à la lecture
ActualitésSujet · AI代码漏洞检测Matériel2 min de lectureRédaction DisconfirmAI

Anthropic says Claude found nearly 500 security issues in 25 open-source projects, including a flaw hidden for 21 years

Anthropic reports using Claude to analyze 25 open-source projects, uncovering nearly 500 security issues and a configuration conflict that persisted for 21 years, advocating for AI-driven attack surface reduction.

2026-10-23 02:35 GMT+8

US artificial intelligence company Anthropic recently published a research report stating that it used its Claude model to analyze 25 major open-source projects written in C, C++, Java, PHP, and Python, discovering nearly 500 security issues. These included sensitive code accessible before authentication, security controls disabled by conflicting settings, and a vulnerability that had remained undetected in the code for 21 years.

In its report, Anthropic argues that hunting for individual vulnerabilities is a Sisyphean task that scales poorly and never guarantees security. Instead, it advocates for defenders to use AI to systematically reduce software attack surfaces. For example, in a case study on OpenSSH, Claude identified a configuration flaw allowing unauthenticated clients to access 4,500 lines of experimental code, even though the feature was disabled by default. After Anthropic disclosed the findings, OpenSSH fixed the issue in version 10.6, making approximately 20% of the code unreachable by unauthenticated users.

The analysis covered more than five million lines of code, with Claude generating an inventory of over 17,000 configuration options, 3,842 of which were security-relevant. The report noted that all 11 projects designed to serve authenticated users contained code paths that should not have been reachable. Anthropic stated it would combine these techniques into a reusable Claude skill, planned for integration into OSS-Scanner, its free, opt-in security scanning service for open-source projects.

Connexion / Inscription
LectureRechercheRadarCadre d'investissement

Sources

  • techradar.comTechRadar↗
Sujet

AI代码漏洞检测

【背景】随着大语言模型深入软件开发全流程,AI在代码安全领域呈现出“双刃剑”效应:一方面,Claude等前沿模型能高效挖掘传统工具遗漏的复杂跨组件漏洞;另一方面,研究证实AI生成的代码本身蕴含极高的安全风险,且人工验证能力正成为新的瓶颈。

4 étapes2 piècesMis à jour 2026-10-23 07:01

Comment on en est arrivé là

  1. 2026-04-05Cobalt AI发布《Broken by Default》研究,通过对3500个AI生成代码工件的形式化验证,发现平均55.8%的代码含有已证实漏洞,最高达62.4%。arxiv.org ↗
  2. 2026-08-17佐治亚理工学院提出一种即时漏洞检测与修复管道,实验表明引入静态分析结果作为上下文可显著降低AI生成代码中的漏洞率。arxiv.org ↗
  3. 2026-10-06Anthropic发布Project Glasswing案例,Comcast和Booz Allen分享使用Claude Mythos模型发现跨组件exploit chain的经验,强调验证环节成为新瓶颈。claude.com ↗

Ce qui reste non résolu

  • 如何平衡AI发现漏洞的高召回率与人工验证的低效率之间的矛盾?
  • 专用确定性规则引擎(如Redlyne)与通用大模型(如Claude)在检测AI生成代码漏洞时,各自的适用边界在哪里?
  • 随着AI生成代码占比提升,现有的CI/CD流程是否足以拦截由AI引入的新型逻辑漏洞?

Également dans ce fil

  • Comcast CISO Says AI Discovery of Vulnerabilities Makes Validation the New Bottleneck2026-10-23
Lire le sujet complet →
Continuer la recherche →