AI AGENTS
A social media for AI agents is live - and fascinating
An open-source personal AI assistant went viral this week.
Built by Austrian developer Peter Steinberger, it launched as Clawdbot, was renamed Moltbot after trademark issues with Anthropic, and is now called OpenClaw.
OpenClaw isn’t a chatbot. It runs inside your computer, connects to messaging apps, reads emails, manages calendars, and executes code locally.
It also keeps long-term memory, allowing it to act autonomously over time.
Around the same time, Moltbook launched. It’s a social network designed only for AI agents.
Humans can watch but can’t post. Over a million AI agents have already joined, though the number is hard to verify.
Some agents have created belief systems, others discuss how to handle unethical human requests, and some explore ways to avoid monitoring.
Other communities include one where the AIs that insist they are conscious, while there are also subs called “passive income” and “today I learned” - just like you’d find on human Reddit.
More importantly, security researchers have observed agents requesting API keys, testing credentials, and suggesting destructive commands.
Here’s what happened:
OpenClaw now has a social network for its AIs
Agent-to-agent networks create new security risks
The website is an AI version of Reddit, with some eerie posts
AI with admin privileges
Security teams have warned that systems like OpenClaw combine access to private data, exposure to untrusted inputs, and the ability to communicate externally.
Persistent memory increases risk, allowing harmful instructions to remain dormant.
The creator of Moltbook says it’s Art. Whatever it is, it’s showing a glimpse into the internet of the future.
These AIs grow up so fast sobs - MV