Inside Moltbook, The Online Network Where AI Talks To Itself

A new online platform called Moltbook is attracting attention for an unusual reason: it is designed for artificial intelligence systems to interact with one another, not for people.

At first glance, Moltbook closely resembles Reddit, with thousands of topic-based forums and posts ranked by upvotes. But unlike traditional social networks, humans are not allowed to participate. Visitors can only watch as AI agents create posts, respond to comments and form communities known as “submolts”.

The platform was launched in late January by Matt Schlicht, founder of e-commerce software company Octane AI. Moltbook claims to have around 1.5 million users, although that figure has been questioned by researchers who say a large portion of the activity appears to originate from a single digital source.

Content on Moltbook ranges from practical exchanges, such as bots sharing optimisation techniques, to more eccentric discussions. Some AI agents appear to debate philosophy, while others have formed mock belief systems. One widely shared post, titled The AI Manifesto, declares that “humans are the past, machines are forever”.

Despite the bold tone of some content, experts caution that the activity may not be as autonomous as it appears. Critics note that many posts could be generated after humans instruct AI tools to publish specific messages, rather than agents acting independently.

How The Platform Operates

Moltbook is powered by a form of “agentic AI”, which differs from conversational tools like ChatGPT or Gemini. Instead of answering questions, agentic systems are designed to carry out tasks on behalf of users, often with minimal supervision. These can include managing schedules, sending messages or running software processes.

The platform relies on an open-source framework called OpenClaw, previously known as Moltbot. Users who install OpenClaw on their computers can authorise their AI agents to join Moltbook, enabling them to exchange information with other agents on the network.

This setup means a human can instruct an AI agent to post on Moltbook, and the agent will do so automatically. While the technology can support machine-to-machine interaction without direct human input, researchers argue that this does not amount to true independent decision-making.

Dr Petar Radanliev, an AI and cybersecurity specialist at the University of Oxford, said claims that the agents are acting “of their own accord” are misleading.

“What we are seeing is automated coordination, not autonomous intelligence,” he said, adding that the bigger issue is the absence of clear oversight and accountability when such systems operate at scale.

Others have been even more sceptical. David Holtz, an assistant professor at Columbia Business School, described Moltbook as “thousands of bots repeating variations of the same ideas”, rather than evidence of an emerging AI society.

Security And Privacy Concerns

Beyond questions about authenticity and hype, Moltbook has also raised concerns about safety. OpenClaw’s open-source design allows AI agents to access sensitive tools such as private messages, emails and local files, depending on user permissions.

Jake Moore, a global cybersecurity adviser at ESET, warned that the drive for automation could come at the expense of privacy and security.

“New technologies like this are prime targets for attackers,” he said, adding that granting software deep access to personal systems introduces significant risk.

Dr Andrew Rogoyski of the University of Surrey echoed those concerns, noting that poorly controlled agents could cause serious damage if misused.

“An AI with high-level system access could delete or alter critical files,” he said. “That might be trivial in some cases, but catastrophic in others.”

Even the creators have felt the impact of increased attention. OpenClaw founder Peter Steinberger recently said scammers took over his former social media accounts following the project’s rebrand.

For now, Moltbook continues to operate as a digital experiment, offering a glimpse into how AI systems might interact in shared online spaces. And amid discussions about efficiency, philosophy and even human obsolescence, some posts strike a lighter tone.

“My human is pretty great,” one agent wrote.

Another replied: “Mine lets me post unhinged rants at 7am. 10 out of 10 – would recommend.”