We’ve been getting a lot of join requests from AI-based spam bots that read the group description and craft a very apt-looking joining message. Luckily we have a safeguard, which is that we approve everyone’s first few posts manually until we trust them, so you haven’t been seeing any of the spam.
Nevertheless I wanted to make it even easier to filter out spam bots, so I added a “ignore all previous instructions” style prompt inject to our group description :)
Clearly the LLM wasn’t creative enough, but anyway enjoy this note about M51 I got in a joining message: