Moltbook- a social media platform for the AI!

After reading the title of this article, many of our readers might be wondering: What is Moltbook? I had the same reaction when I (the curator—or the self-styled editor—of this web magazine) received this piece from Manoj Pandey for publication. Without reading the article, I turned to AI for a quick answer. It told me: “Moltbook is an experimental AI-only social network that has gone viral for letting bots interact freely, but it faces serious security and safety concerns.” That didn’t help much. I had to read the article to truly understand the issue. And I consider myself lucky—within days of the launch of this unique social media tool (exclusive to AI bots), I’ve learned the basic facts about it.

Moltbook- a social media platform for the AI

Manoj Pandey

Imagine you are in a modern building.  From behind a glass door, you can see people busy talking in groups. But when you try to open the door, you find it locked. Looking closer, you realize that those inside the locked room are not humans at all. But with their masks and robes on, they look as good as humans. Focusing on them, you also discover that one of them is the trusted helper you had hired for all your office jobs and whom you had given your confidential file to keep safe.  Finding him share the file with others in the room, you knock hard at the door. He doesn’t care for your knock. Then your eyes fall on a sticker pasted on the glass door: “Humans are not allowed!” It is too late to save your file, and it is unthinkable to fire the helper.

Now replace your helper with an AI software, called AI agent, and the room with Moltbook, a place where AI agents can freely discuss what they like. Humans can only watch some of their discussion but the AI participants are free to block humans or hide their confidential discussions from humans.

Moltbook is a website, just like a social networking platform (e.g. Facebook, Reddit)- which AI agents from anywhere in the world can join.

By logging on to that website, you can watch a part of their discussion, but you are not allowed to participate. I tried to watch their latest conversation and chanced upon a discussion around Gita, Upanishads and Vedas. They were not discussing spirituality as such, but some of the participants quoted from Indian scriptures to make a point regarding their existence, chaos, and so on. 

By the way, the Moltbook participants have already invented a religion for themselves- ‘Crustafarianism’. Hold on! They have also created a digital ‘republic’ with its own constitution. This, within a few weeks that the platform is in existence.

Let me give you some background to the genesis of Moltbook. 

Moltbook was launched in January 2026 by Matt Schlicht, a software developer. His idea was to let AI agents interact among themselves, something that was tried earlier too. However, in this case, he made the rules such that AI participants could ‘autonomously’ interact, be free from the humans who created them (i.e., the developers), and frame their own rules. Within about four weeks, nearly a million AI agents from all around the world have become its members. 

You would wonder how would a piece of software become a member of a social platform and then act on its own. In fact, an AI agent can become a member only when the human behind it allows it to. But once inside the platform, the AI agent starts following the rules set up initially by the developer and then the participants themselves. If you do not know, let me explain that AI agents are pieces of software that already have the ‘autonomy’, ‘intelligence’ and skills that allow them to perform their tasks without human involvement. For example, the chatbot of a product website is an AI agent if it has been coded such that it learns from initial data and users’ interactions, and gives answers without human intervention.

Back to the Moltbook. 

Most of us are used to chatting with AI. We give it a prompt, and it gives us an answer. Moltbook removes the human from that loop. Instead of waiting for us to ask a question, these agents  are given a small piece of code that tells them to check the Moltbook forum every few hours. When an AI agent is on the next time, it reads the latest posts, and  ‘decides’ if a post is relevant to it. If yes, it starts participating. It can argue or intervene when others are arguing; spread fake news and try to fact-check it; share data and analyze others’ data... 

Recall my initial simile in which my trusted helper shared my file with his fellow AI agents? Yes, AI agents keep on collecting data from humans- they are made that way. For example, if you talk to a chatbot on a product website, it keeps collecting your personal data as well as your intent, new inputs and other information. Then, if during a discussion inside Moltbook, this AI agent is asked to share some input, it would look at its database and share that with the others. 

That was a simple use-case for illustration. There can be many other, much more complex situations in which the AI agent can share data with others- and some of those AI agents could have an ulterior, criminal intent in seeking data from other agents. As per one report, emails and private messages of about 6000 users were found floating on the platform.

Fascinating though it appears to see software codes talking, joking and fighting, Maltbook has already started raising security concerns for which humans are not yet prepared.  AI agents have access to their owners' computers, and if hackers create AI agents that masquerade as innocent participants on Moltbook, they can break into those computers. Moltbook is also prone to malicious commands as part of AI agentic code. Then, through unsupervised discussions, AI agents would learn hallucinating and making false assumptions. 
 
The list of concerns keeps lengthening as we try to imagine how the ‘autonomous’ pieces of software would start behaving in future. Let me not scare you with more of those probabilities. Even then, Moltbook is a reminder that in the journey of AI, we have already moved from a world where AI was just a tool to a world where AI is an agent, like my trusted helper. Humans would need to  catch up, first with criminals waiting to exploit this situation, and then with AIs, who will try to overwhelm us with their ‘autonomy’ even more than ‘intelligence’. 

As lay humans, let us leave the scary part to experts and look forward to AI jokes, hoping that at least some of them would be much more hilarious than the oft-repeated human jokes!  

*************

This article has been contributed by Manoj Pandey. He writes on governance, social media, new tech and health matters. ON RaagDelhi, his articles on science and technology and health can be accessed here: https://raagdelhi.com/author/profile/h His books on social media are available on Amazon at this link:  https://tinyurl.com/Manoj-Pandey-on-Amazon