The US National Security Agency is reportedly using Anthropic’s cutting-edge Mythos AI model for hacking.
Mythos has not been publicly released because of its powerful abilities to detect and exploit software vulnerabilities, and instead is being offered to select institutions to help them build defenses.
The NSA is apparently using it for offensive purposes even as the US government fights a legal battle against Anthropic.
AI agents are increasingly ubiquitous on the web — Cloudflare said that most traffic is now AI-led, a year ahead of earlier projections, with a decent minority of them malicious bots; Iran is a hotspot for such activity. And while such bots lack Mythos’ cutting-edge capabilities, top open-source models are less than a year behind the frontier systems.