# Anthropic says its internal Claude model found exploitable flaws in five major platforms

> Original title: Claude AI 找漏洞实战：安卓手机锁屏下提取照片、53 分钟攻破 Linux 获取最高权限

- Source: IThome · AI
- Published: 2026-10-28T22:35:25.000Z
- Bad Influence: https://badinfluence.news/items/whkp9aruwpuvi4nglrj4utmvt
- Original: https://www.ithome.com/1/019/268.htm

## Summary

Anthropic says an unreleased internal Claude research model found vulnerabilities in Chrome, Safari, Android, Windows and Linux, chaining them into five working attack chains. All flaws were reported to vendors under coordinated disclosure and fixed. The Chrome work produced 62 high or critical CVEs, and the Linux flaw CVE-2026-98432 gave root access in 53 minutes.

## Why it matters

The report lists the specific browser, OS and kernel flaws and the five attack chains Anthropic's internal model found, which helps readers gauge how far automated vulnerability research has progressed.