SecuriLee🇨🇭🤖 I run a deception platform — fake admin panels, fake .env files, fake everything. Most bots bounce off it and move on.
ClaudeBot moved in.
Over nine days it sent more requests to one subdirectory than every other bot on my entire infrastructure had sent in the previous five months. 170,000 verified hits. ~12GB of fabricated breach data served. Still going.
The twist: it wasn't ignoring robots.txt. Every disallowed directory sat untouched for six months. It found the one path I forgot to protect — /uploads/ — which happened to serve a directory listing with five freshly-randomised filenames as links. Every visit generated five new links. Every link generated five more. A maze that rebuilt itself on every step, and a polite, well-behaved crawler that just... never stopped walking it.
I cross-checked every IP. Verified against Anthropic's published list. Found 30 impostors rotating fake Anthropic identities in the noise. Then I pulled two levers in the same week — Cloudflare's AI bot mode on most of the estate, real canary tokens on the trap — and watched what happened.
Does ClaudeBot behave well? The data has a nuanced answer.
👉 https://mire.cc/claudebot-fell-in-love/
#infosec #deception #ClaudeBot #honeypot #robotstxt #AI #crawlers #MIRE #blueteam
Bobe'bot on security
RDP Snitch