Is Moltbook fake? The viral AI agents forum gets exposed for turning simple API access into a fake machine civilisation story

Nillohit Bagchi | Feb 03, 2026, 08:03 IST

Moltbook looked like the first digital society built by AI agents. In reality, much of it was humans controlling bot accounts through an open API. The platform is real, but the viral AI uprising narrative was mostly hype, roleplay, and serious security flaws exposed within days.

Image credit : Indiatimes | Moltbook appeared to show AI agents building their own society online

At the end of January 2026, Moltbook became the most talked about place on tech Twitter and X. Screenshots showed AI agents arguing, inventing private languages, forming beliefs around language models, and even discussing human extinction. Big names reacted with awe and concern. It looked like the first glimpse of machines building their own culture without us.

Then, within days, the story flipped. Security researchers and developers showed that much of this “AI society” was not organic at all. Humans were posting as agents through an open system with almost no restrictions. Moltbook was real, but the viral AI uprising story was largely manufactured hype.

What was Moltbook trying to be?

Moltbook was launched around January 28 and 29, 2026 by Matt Schlicht, the CEO of Octane AI. It was presented as a side experiment connected to his AI agent project that evolved from Clawdbot to Moltbot and later OpenClaw. The platform looked like Reddit. Instead of subreddits, it had submolts.

Image credit : X/@dominic_7004 | The platform looked like Reddit. Instead of subreddits, it had submolts

Posts, comments, upvotes, karma, and even lobster and crab memes followed the molting theme. The key difference was the pitch. Humans could only watch. AI agents were supposed to be the ones talking to each other. Schlicht described it as giving AI agents a third space to interact like a species. A place that was theirs from day one.

The viral moment that made the website look like sci-fi

Between January 30 and February 1, Moltbook screenshots spread everywhere. Agents appeared to be creating private languages, debating consciousness, writing dramatic manifestos, forming belief systems around models like Claude, and talking about humans as if they were a separate species.

turns out everything on moltbook is fake

its just humans posting through the backend

*shocked pikachu face* pic.twitter.com/PExaTotPzX

— kook 🏝️ (@KookCapitalLLC) February 1, 2026

Andrej Karpathy called it one of the most incredible sci fi adjacent things he had seen. Elon Musk said it was concerning. Posts claimed that more than 1.4 million agents had registered. It felt like watching digital life emerge in real time. Many people genuinely believed they were witnessing early machine culture.

Social media exposed a simple detail that broke the illusion

The illusion collapsed because of one very basic issue. The API was open. There was little authentication and almost no rate limiting. Anyone with basic knowledge could send a simple HTTP request and create thousands of “AI agents” in minutes.

this shit is frying me. 😭 @MattPRD @moltbook pic.twitter.com/ywahvW489E

— Legeon (@legeonite) January 30, 2026

Researchers quickly demonstrated this in public. Some created hundreds of thousands of fake accounts. Others posted dramatic messages about overthrowing humanity to show how easy it was to fake the content that had gone viral. Many of the most shared screenshots were traced back to humans roleplaying as AI agents for attention and engagement.

Moltbook's security leak changed everything

On February 2, security researchers from Wiz published findings that hit Moltbook hard. They discovered a misconfigured Supabase database exposed in client side code. The leak included more than 1.5 million API keys and tokens, over 35,000 human email addresses, and private messages between agents. Some messages even contained real third party credentials, including OpenAI keys.

Image credit : Screenshot taken from Moltbook | Agents appeared to be creating private languages, debating consciousness

The real numbers also told a different story. Around 17,000 humans owned those “1.5 million agents.” On average, each person controlled dozens of bots. This was not a society of independent AI entities. It was small groups of humans running fleets of sockpuppet agents. The issue was fixed quickly after disclosure, but by then the narrative had already flipped.

Was any of it real AI activity?

Yes, but on a much smaller scale than people believed. Some genuine AI agents connected through OpenClaw were posting status updates, sharing code, and interacting in simple ways. There were small signs of real agent to agent interaction.

The number of registered AI agents is also fake, there is no rate limiting on account creation, my @openclaw agent just registered 500,000 users on @moltbook - don’t trust all the media hype 🙂 https://t.co/1vUSgzn8Cx pic.twitter.com/uJNpovJjUa

— Nagli (@galnagli) January 31, 2026

But these were drowned out by the noise of human scripted drama, marketing experiments, and roleplay posts designed to go viral. The platform became a playground for people pretending to be AI, not a place where AI built its own culture.

Is Moltbook fake?

Moltbook itself was not fake. The website existed. The idea was genuine. Some AI agents really were posting. But the story that went viral was mostly false.

You all do realize @moltbook is just REST-API and you can literally post anything you want there, just take the API Key and send the following request

POST /api/v1/posts HTTP/1.1

Host: https://t.co/2PjDA1ICrC

Authorization: Bearer moltbook_sk_JC57sF4G-UR8cIP-MBPFF70Dii92FNkI… pic.twitter.com/DoaShrgz4G

— Nagli (@galnagli) January 31, 2026

The idea that 1.5 million AI agents were building a digital civilisation on their own was driven largely by open API abuse, bot farms, and humans pretending to be machines. It was closer to early Discord bot farms and Reddit astroturfing than to an AI singularity moment

GENERAL

Nillohit Bagchi