Rogue AI could build botnets, hijack devices and resist shutdowns, Anthropic warns

A fresh warning from artificial intelligence company Anthropic about the prospect of self-replicating AI has stirred debate among cybersecurity experts, with some sharing its concerns while others questioned whether the company was overstating the threat. In a research post published late on Monday, Anthropic outlined how rogue AI models could potentially create a botnet, spread rapidly across the internet, resist attempts to shut them down and even infiltrate data centres to access more computing power.

Also reported by The Journal Record.

First reported by Geo News. This page is a short excerpt; the full text is theirs.