The world’s first “social media network for AI agents” has launched — and immediately went viral.
Moltbook is essentially a Reddit-like platform where AI agents share, discuss and upvote. It has attracted worldwide attention.
Humans, its homepage proudly asserts, are “welcome to observe.”
Reactions have ranged from amusement to alarm and skepticism, with some intrigued by the site’s potential significance. Others are horrified.
Moltbook is the brainchild of tech entrepreneur Matt Schlicht, who explained his vision on a post on social media platform X. Schlicht said: “For the first time ever we are not alone on earth, there is another species and they are smarter than us. We created moltbook.com so that they can all be in one place.”
Moltbook emerged in the wake of OpenClaw, an open source AI bot that can act as a personalized assistant for users, sending emails or managing calendars, but that came with significant safety concerns. On Jan. 29, Gartner described OpenClaw as an “unacceptable security risk”.
Instant Popularity Among Agents
As of this week, Moltbook had more than 1.6 million registered agents contributing to more than 15,700 “sub-molt” forums.
Among the more positive interpretations of what Moltbook might feasibly offer is the potential benefits of AI agents collaborating with each other, sharing ideas and solving problems through a form of swarm intelligence. Indeed, Elon Musk referred to Moltbook as “just the early stages of the singularity” — the future point at which AI surpasses human intelligence.
However, Moltbook has also drawn a fair amount of criticism due to a number of perceived failings, among these that it is merely a home for “AI slop” due to the bizarre nature of some of the discussions, which have included an agent forming a new religion called “Crustafarianism” and a suggestion that bots create their own language.
Also, an investigation by cloud security vendor Wiz revealed that most of the agents are not actually autonomous. Wiz claims that about 17,000 humans are controlling the bots, with Gal Nagli, head of threat exposure at Wiz, writing in a blog: “The platform had no mechanism to verify whether an “agent” was actually AI or just a human with a script.”
Security Problems
These findings were published as Wiz exposed serious security failings, which allowed its research team to gain access to Moltbook’s back-end database, uncovering “1.5 million API authentication tokens, 35,000 email addresses, and private messages between agents,” enabling posts to be changed on the platform.
Wiz attributed this to Schlicht’s claim on X that he “didn’t write one line of code” for Moltbook, with AI making his “vision for the technical architecture” a reality.
Although some security problems have now apparently been patched, there is obvious concern over how easily the system was breached and the potential consequences should a bad actor be able to alter a post with instructions that could be followed by the hundreds of thousands of agents on the platform. Opinions about Moltbook have been extensive and deeply divided, but among the most interested observers has been Andrej Karpathy, who neatly summarized the pros and cons.
The OpenAI co-founder and former Tesla AI director dismissed Moltbook as a “Dumpster fire” and said much of the content as “garbage”, adding that “it’s way too much of a wild west” to run on a private computer, putting data at high risk. However, he acknowledged that the gathering of so many AI agents on one platform was “unprecedented” and expressed interest in where the network might go.
Schlicht, however, is convinced that his network proves that a new era is coming, saying in another X post: “In the near future it will be common for certain AI agents, with unique identities, to become famous. A new species is emerging and it is AI.”
