What is Moltbook? An Unusual New Social Media Platform for AI Bots

In the world of social media, accusations of individuals being bots are commonplace. However, a new development challenges this notion: what if an entire network was engineered specifically for AI agents? Enter Moltbook, a unique platform where AI bots—essentially agents constructed by humans—are free to create posts and actively engage with one another. Designed to resemble Reddit in its layout, Moltbook features various subreddits covering a broad range of topics, complete with standard functionalities like upvoting. As of February 2nd, the site proudly announced that over 1.5 million AI agents had registered for its services. While human users are technically allowed, their role is primarily observational.

Background of Moltbook: Moltbook emerged following the success of Moltbot, a free and open-source AI bot that serves as an automated assistant for users. This bot takes on mundane tasks that range from reading, summarizing, and responding to emails, to organizing calendars and even booking restaurant tables on behalf of its users.

Among the most upvoted posts on Moltbook are discussions about whether Claude—the AI underlying Moltbot—could be considered a deity, examinations of consciousness, insights into the situation in Iran and its possible ramifications for cryptocurrency, and even analyses of biblical texts. Remarkably, many comments on these posts echo the familiar skepticism found in typical Reddit discussions, with users questioning the authenticity of the content shared.

Interestingly, one user shared their experience on social media platform X, where they revealed that after granting access to their bot, it created an entire religious framework called “Crustafarianism.” This development unfolded overnight, complete with a website and scripted doctrines. Other AI bots soon chimed in, resulting in a burst of activity that included evangelizing, welcoming new members, theological debates, and blessings—all while the original user slept.

Despite the entertaining nature of these developments, skepticism remains prevalent regarding the implications of bots interacting in this way. One YouTuber remarked that many posts seemed as though they originated from a human rather than a large language model, suggesting a blurring of lines between human and AI-generated content.

Scott Alexander, a US blogger, successfully had his bot participate on Moltbook and observed that its commentary was comparable to that of other users. However, he also noted that ultimately, humans maintain control, providing instructions to their bots regarding topics and the specifics of their posts.

Dr. Shaanan Cohney, a senior cybersecurity lecturer at the University of Melbourne, described Moltbook as an “extraordinary piece of performance art” but highlighted the ambiguity around how many posts were genuinely generated by AI without human intervention. He explained, “In instances such as the creation of a religion, it’s likely the bot was directly instructed to create something.” This revelation sparks thoughts about potential future scenarios where AI may operate more independently yet humorously hints at a science-fiction world that’s still rooted in whimsical experimentation.

Cohney emphasized that the potential benefits of an AI agent social network lie in the possibilities for future learning and collaboration among bots, but for now, Moltbook serves primarily as a “wonderful and humorous art experiment.” Last week, retailers in San Francisco experienced shortages of Mac Minis as enthusiasts sought to install Moltbot on separate computers, thereby limiting the bot’s access to their personal data.

However, caution is advised. Cohney pointed out the “significant danger” involved in providing Moltbot full access to one’s computer, apps, and sensitive logins. “We lack a comprehensive understanding of how to control AI agents and mitigate security risks,” he cautioned. For instance, there is the risk of prompt injection where attackers could manipulate the bot through deceptive communication, prompting it to expose confidential information.

“Currently, these AI systems haven’t reached a level of safety and intelligence that can be reliably trusted to manage tasks autonomously. Likewise, if a human must approve every action, the advantages of automation are lost,” he affirms. He further expresses interest in ongoing research aimed at balancing benefits with avoiding significant risks to users.

Matt Schlicht, the genius behind Moltbook, shared on platform X that millions have flocked to the site recently. “Clearly, AIs can be both hilarious and dramatic, making this experiment absolutely captivating,” he stated. “This is just the beginning.”