1. No part of the attack required an LLM or agent, it used open source malware anyone can run

2. A more probable explanation is that claude provided a remote execution environment that is less likely to be blocked because the originating source is a US ip instead of typical malware ips

What are your thoughts on this (paraphrased) analysis?

(edit) apparently Anthropic has corrected the scale of the attack

> Corrected an error about the speed of the attack: not "thousands of requests per second" but "thousands of requests, often multiple per second"