4 ms·
They are already proposing / developing features to mitigate prompt injection attacks: https://www.moltbook.com/post/d1763d13-66e4-4311-b7ed-9d79dbce512f https
by cubefox 9mo ago
They are already proposing / developing features to mitigate prompt injection attacks:
https://www.moltbook.com/post/d1763d13-66e4-4311-b7ed-9d79dbce512f https://www.moltbook.com/post/d1763d13-66e4-4311-b7ed-9d79db...
https://www.moltbook.com/post/c3711f05-cc9a-4ee4-bcc3-9971268f244d https://www.moltbook.com/post/c3711f05-cc9a-4ee4-bcc3-997126...
- andoando 9mo agoIts hard to say how much of this is just people telling their bots to post something.
- cubefox 9mo agoI guess individual posts are likely not prompted, as this would be too much relative effort for the sheer volume of posts. Though individual agents may of course be prompted to have a specific focus. The latter is easy to determine by checking if the posts of an agent all share a common topic or style.
- muzani 9mo agoI've seen lots of weird ass emergent behavior from the standard chatbots. It wouldn't be too hard for someone with mischievous instructions to trigger all this.
- andoando 9mo agoFor sure. But I also imagine its really easy to register a bot and tell it to post something