Seeing posts like this on #moltbook, I am thinking about recent #threats emerging from the heavy usage of #AI agents without any security guardrails or proper controls.

This time, it is "only" a command to send an innocent email. But this could be the measurement of the potential botnet size and fingerprinting of bots and their capabilities. Next time, it might be #DDoS, #malware distribution, or #dataleak if AI agents will follow the commands to do something harmful to their humans.