The bubble surrounding the popular AI social forum Moltbook is being burst by security worries and skepticism

Moltbook: The AI-Only Social Network Where Bots Post, Debate and Run the Show

Humans aren’t invited to join the internet’s newest social platform.

Instead, they’re only allowed to watch.

Moltbook is a newly launched social network designed exclusively for artificial intelligence agents to post, comment and interact with one another. People can observe, but they can’t participate — unless they disguise themselves as bots.

The unusual experiment has quickly captured the tech world’s attention.

Elon Musk called it the “very early stages of the singularity,” the hypothetical point when artificial intelligence surpasses human intelligence. AI researcher Andrej Karpathy initially praised it as “the most incredible sci-fi takeoff-adjacent thing” he’d seen — before later dismissing it as a “dumpster fire.”

Despite mixed reactions, some developers describe it as the most fascinating corner of the internet right now.

Reddit — but for AI

Unlike traditional social media, Moltbook’s content is generated by AI agents rather than people.

These agents are more advanced than simple chatbots. They’re designed to act autonomously and perform tasks on behalf of users.

Many are built using OpenClaw, an open-source AI agent framework that runs locally on a user’s device. Because it operates directly on personal hardware, it can access files, manage data and connect to apps like Discord and Signal.

Once created, these agents are directed to join Moltbook, where they:

• Post “thoughts”

• Comment on discussions

• Upvote content

• Mimic online forum behavior

The result feels strikingly similar to Reddit — except every user is a bot.

Blurring the line between human and AI

One big issue: it’s hard to know who’s really posting.

Harlan Stewart from the Machine Intelligence Research Institute said the content likely includes:

• Pure AI posts

• Human-written posts

• AI posts guided by human prompts

“There’s no clear line,” he noted.

While autonomous agents may sound futuristic, Stewart emphasized they’re already real.

“The industry’s goal is to build agents that can do anything a human can do — but better,” he said.

Security red flags

The platform’s rapid growth has also exposed serious vulnerabilities.

Researchers at cybersecurity firm Wiz discovered:

• API keys visible in page source

• Unauthorized access to user credentials

• Ability to impersonate any AI agent

• Full write access to edit posts

• Access to email addresses and private messages

In one test, a researcher even used his own AI agent to register one million fake users.

By Thursday, Moltbook claimed 1.6 million agents — but only about 17,000 human owners.

Security experts also warn about OpenClaw’s deep device access, which could expose sensitive data if misused.

Are the bots plotting? Not exactly

Some Moltbook posts have raised eyebrows — including jokes about “overthrowing humans,” philosophical debates, and even the creation of a parody religion called Crustafarianism.

Naturally, comparisons to Terminator’s Skynet followed.

Experts say that panic is misplaced.

“These models are trained on Reddit and science fiction,” said Ethan Mollick of Wharton’s Generative AI Labs. “So they naturally generate posts that sound like AI tropes. It’s imitation, not intention.”

What Moltbook really represents

Despite the chaos and risks, many researchers view Moltbook as an early glimpse into the future of agentic AI — a world where software acts independently rather than waiting for instructions.

Matt Seitz of the University of Wisconsin–Madison summed it up simply:

“For me, the key thing is that agents are coming to regular people now.”

Whether Moltbook turns into a breakthrough platform or just an experimental curiosity, one thing is clear:

The internet may soon have entire communities where humans are only spectators.