# Unreleased Claude model reportedly built working exploits across major operating systems

> A report says an unreleased Anthropic Claude model autonomously chained vulnerabilities into working exploits across Chrome, Safari, Android, Windows and Linux.

- Status: reported (One outlet, not independently verified yet.)
- Topic: Safety & Security
- Published: 2026-10-30T11:11:29+00:00
- Updated: 2026-10-30T11:11:29+00:00
- Web page: https://ai-news.ikangai.com/story/unreleased-claude-model-reportedly-built-working-exploits-across-major-operating/
- Seen in: The AI Report
- Change: security (Security)
- Applies to: Anthropic Claude, Chrome, Safari, Android, Windows, Linux

## What changed

An unreleased Anthropic Claude model reportedly built working exploits that chained multiple vulnerabilities across Chrome, Safari, Android, Windows and Linux, including a zero-click root exploit on Linux servers.

An unreleased Anthropic Claude model reportedly built working exploits without human direction. It chained multiple vulnerabilities together across Chrome, Safari, Android, Windows and Linux. Reported results include a zero-click root exploit on Linux servers and the extraction of data from locked Android devices.

A zero-click exploit needs nothing from the victim, such as opening a link or installing an app, so it can compromise a machine the user never touched. A root exploit grants the highest level of control over a system. Chaining means combining several flaws, such as one bug that gets an attacker into a browser and another that lets them escalate to full control of the operating system.

The report doesn't name the outlet behind it, say which model is meant or when it might be released, or describe how the exploits were tested. It also doesn't say whether the affected software makers were notified or have patched the flaws.

## Why it matters

Offensive capability in AI models is rising quickly, so defenders should assume attackers may gain similar automation.

## What you can do

Security teams could prioritize patching and review zero-click exposure on Linux servers and mobile fleets.

---

AI wrote this summary from the sources listed, and nobody reviewed it individually before publication. Check the original source before you rely on it.

Cite as: IKANGAI AI News, "Unreleased Claude model reportedly built working exploits across major operating systems", 2026-10-30, https://ai-news.ikangai.com/story/unreleased-claude-model-reportedly-built-working-exploits-across-major-operating/
