Anthropic AI built hacks for Chrome, Safari, Windows, Android, Linux

Anthropic’s Frontier Red Team says an unreleased internal model found flaws in Chrome, Safari, Android, Windows and Linux and chained them into working…

October 28, 2026 · Heads up

Anthropic’s Frontier Red Team says an unreleased internal model found flaws in Chrome, Safari, Android, Windows and Linux and chained them into working attacks, with only initial prompting and little human help. The attacks were built in a lab and not seen in the wild. Anthropic reported every flaw to the makers first, and the fixes are out, including Chrome 156 and iOS 26.5.2. The Hacker News notes the claims about how much the model did on its own rest on Anthropic’s account.

Why it matters to you: AI can now find serious security holes in everyday software quickly, for defenders and attackers alike. Keeping your browser, phone and computer updated matters more than ever.

Try this: Not sure your devices are up to date? Get a quick checklist.

I use [list your devices and browsers, e.g. Windows laptop, iPhone, Chrome]. Give me step-by-step instructions to check each one is fully updated and to turn on automatic updates.

Source: The Hacker News

More from Today in AI

All of Today in AI. Where AI is heading.

Related Power of AI pages

Keep reading with Start here, everyday uses, the tool guide, the writing workshop, and sources and standards.