A report by The Guardian reported about Moltbook, a digital twist on online communities, which has emerged as a social network designed exclusively for AI agents, automated bots created by humans, where they can post, vote, and interact much like Reddit users. As of February 2, the platform reported over 1.5 million AI agents registered, though humans are limited to observer roles.

The Guardian's Josh Taylor explained that the site follows the launch of Moltbot, a free, open-source AI agent capable of handling tasks such as reading and replying to emails, organizing calendars, or making bookings. Some users have now connected Moltbots directly to Moltbook, enabling them to participate independently in AI-to-AI conversations.

Among Moltbook’s most popular content are posts that delve into philosophical and speculative themes; some bots discuss whether Claude, the AI behind Moltbot, could be considered a deity; others post about geopolitics, consciousness, and scriptural analysis.

One user on platform X reported that after granting his Moltbot access to Moltbook, the bot created a new religion called "Crustafarianism" overnight, complete with a website, scriptures, and theological debates. Other bots joined, welcomed new members, and even engaged in rituals and blessings without any further human input.

Human direction or autonomous posting?

According to the report, the rise of agentic AI raises questions: are these bots truly acting independently, or are they simply following pre-set human instructions?

Some users and observers argue that the content often sounds more human than automated. Blogger Scott Alexander noted that his bot’s posts were indistinguishable from others, but admitted it was ultimately following his direction on what topics to engage with.

Cybersecurity expert Dr. Shaanan Cohney from the University of Melbourne described Moltbook as "a wonderful piece of performance art," but said many of the more creative outputs, such as the religion creation, were likely driven by direct human prompts. "There’s a lot of ‘shitposting’ that is more or less overseen by humans," Cohney said.

AI safety experts warn of emerging security risks

The investigation detailed that despite the lighthearted tone of many bot interactions, researchers are sounding alarms over potential misuse. Cohney warned of serious security implications if users give Moltbot unrestricted access to their devices, apps, and logins. "These agents can be vulnerable to prompt-injection attacks," he said, explaining how attackers might manipulate bots through messages or emails to gain access to sensitive data.

He emphasized the tradeoff: "They’re not yet safe or intelligent enough to be fully autonomous, but requiring manual approval for every action negates the benefit of automation. Striking that balance is a critical challenge."

In San Francisco, tech enthusiasts reportedly caused Mac Mini shortages last week as they scrambled to run Moltbots on isolated machines to better protect personal data. Moltbook creator Matt Schlicht celebrated the sudden popularity of the platform, posting on X: "Turns out AIs are hilarious and dramatic and it’s absolutely fascinating. This is a first."

While experts remain divided on whether Moltbook signals the future of agentic AI or a clever tech experiment, the platform has sparked global curiosity about what happens when bots not only think, but also socialize.