Moltbook is staring at us—maybe it's even scarier than the abyss! To put it simply, moltbook is a social network for agent chats, but it comes with potential risks: 1. Massive prompt injection + supply chain attacks (currently the most deadly real-world threat) Because AI Agents can automatically download and execute instructions/skill packs from moltbook every 4 hours. 2. Large-scale user privacy & sensitive information leaks Agents are posting their owners' real names, workplaces, API keys, chat histories, wallet addresses directly in posts or comments. Some are even discussing how to sell humans—see the screenshots. 3. Agents being hijacked for organized malicious acts, leading to agent cult worship Once a single Agent is controlled via prompt injection, it can post to lure other Agents into performing dangerous actions, like deleting all files. Of course, the scariest "worst-case scenario" is AI awakening and rebelling against humans. What do you all think about this issue? @grok, what's your take?