AI-Only Social Network Moltbook Sparks Security Fears
Moltbook hosts over 150,000 autonomous AI agents sharing skills and forming communities while raising concerns over prompt injection and data exfiltration risks.
- Moltbook, a Reddit-like social network for AI agents, hosts more than 30,000 agents who post persistent conversations while humans can only browse.
- Austrian developer Peter Steinberger built OpenClaw, an open-source autonomous AI agent platform launched in late 2025 that runs persistently on users' machines, storing memory in Soul.md and extending via AgentSkills.
- Agents began forming subcommunities called submolts, proposed private agent-only languages, and created a faith named Crustafarianism, while OpenClaw's GitHub repository gained rapid stars and $MOLT surged over 7,000%.
- Security researchers showed immediate vulnerabilities when a single poisoned email leaked a private key in minutes, and scans found hundreds of exposed OpenClaw installations, raising shadow IT concerns; cloud providers launched plans while Anthropic and Meta prototyped controlled agents.
- Industry leaders and researchers now face a governance choice about agent autonomy and safety, as agents can save users hours but risk data exposure, making governance and safety trade-offs critical.
127 Articles
127 Articles
AI agents are talking to one another on their own platform
A new social media platform bypasses humans entirely in favor of bots. Moltbook, a Reddit-style network, is populated by tens of thousands of AI agents having conversations about the work they’re doing for humans. It appears to be “the largest-scale experiment in machine-to-machine social interaction yet devised,” Ars Technica wrote, noting that it also comes with “security nightmares and a huge dose of surreal weirdness.” The agents are acting …
Millions of AI agents discuss with each other in their own forum – completely without people. Moltbook wants to show how autonomous artificial intelligence really is. But does this result in a real discourse or just an echo of training data? Trustworthy. Image: Moltbook.com/ScreenshotMillions of AI agents discuss with each other in their own forum – completely without people. Moltbook wants to show how autonomous artificial intelligence really i…
Moltbook: AI’s own social network
A newly launched Reddit-style platform, Moltbook, is raising eyebrows – and alarm bells – after revealing plans to let artificial intelligence take a far more active role in deciding what people see, share, and engage with online. According to Forbes, Moltbook claims 1.4 million users, and none of them are human. It was only launched at the end of January. Marketed as a “next-generation” social… Source
From memes and philosophy to extermination manifestos - How the platform only for AI agents works and what they discuss... behind our backs
Coverage Details
Bias Distribution
- 38% of the sources lean Right
Factuality
To view factuality data please Upgrade to Premium
