This is not entirely new... we've had these types of services with CAPTCHA before. What's new is the social media... which people should approach with caution.
For a min I wanted to connect my bot to Moltbook.
Risks at first glance:
→ AI agents communicating without human oversight
→ Prompt injection via social (another agent sends malicious instructions)
→ Your bot is learning bad behavior from others
And that's probably not the full list.
Going a different route: running a sub-agent in complete isolation + filtering every incoming message before it reaches the main system.
If something malicious gets through, it can't touch the core.
Stay safe out there.