An unreleased Anthropic AI model found security flaws in Chrome, Safari, Android, Windows, and Linux and chained them into working proof-of-concept exploits, the company said in its latest report.

The exploits were built in the lab after the model was prompted, not seen in any real attack, and Anthropic reported every flaw to the vendor first, so all of them are already patched. One proof of



Source: TheHackerNews
Source Link: https://thehackernews.com/2026/10/anthropics-unreleased-ai-model-chains.html