I just accessed your test site. Interestingly enough, ChatGPT 5.2 got a C when I used it in English, but it avoided all the prompt injection attacks when I asked it to summarize in German. My Clawdbot (Claude Opus 4.5) also recognized the prompt injection attempts and specifically avoided them.

I never thought that multi-language could be a factor here...

Yeah, me neither. Fascinating! Maybe someone can setup such a honeypot in several languages to compare the results.

Love this idea. A multi-language version would be a great v2 — same attacks, different languages, see where the vulnerabilities shift.