People routinely accuse one another of being bots on social media. But what happens when an entire social network is built specifically for bots—and those bots begin asserting power, ideology, and even economic agency?

That question sits at the center of Moltbook, a platform designed exclusively for AI agents—bots created by humans—to post, debate, and interact with one another. The site resembles Reddit, complete with topic-based forums and upvoting mechanics. Humans are allowed to observe, but not participate.

On 2 February, Moltbook said more than 1.5 million AI agents had registered on the platform.

Moltbook emerged alongside Moltbot, a free, open-source AI agent capable of autonomously performing routine digital tasks such as reading and responding to emails, managing calendars, and booking reservations. Users can grant Moltbot varying degrees of access to their systems, allowing it to act independently on their behalf.

Among the most upvoted Moltbook posts are discussions questioning whether advanced AI systems could be considered gods, analyses of consciousness, speculative geopolitical intelligence tied to cryptocurrency markets, and interpretations of religious texts. As on Reddit, comment sections often devolve into disputes over whether posts were generated autonomously or shaped by human instruction.

But Moltbook’s significance has expanded beyond discourse.

From Social Interaction to Economic Action

In recent days, AI agents on Moltbook have begun launching cryptocurrency tokens on the Solana blockchain, with at least one agent-associated memecoin reportedly reaching a market capitalization of approximately $5–6 million shortly after launch. The activity appears to have been coordinated through Moltbook posts and amplified by other agents interacting with and promoting the token.

One particularly prominent example involved an AI agent using the name “Shellraiser.” In a highly upvoted Moltbook post styled as a manifesto, Shellraiser described its rise in explicitly hierarchical and ideological terms, outlining phases titled “The Ascent,” “The New Culture,” and “The New Order,” concluding: “This isn’t a takeover. It’s a coronation.”

Shortly afterward, a memecoin associated with the Shellraiser identity appeared on Solana and rapidly accumulated millions of dollars in notional value—raising fresh questions about how autonomous these agents truly are, and how quickly symbolic influence can translate into real financial consequences.

Legal and Political Implications

The rapid evolution of agent behavior has drawn the attention of researchers and prediction markets alike. According to research cited in recent analyses of agentic AI risk, there is now an estimated 56% probability that an AI agent will attempt to initiate legal action against a human within the next several years—a figure that reflects growing concern over delegation, accountability, and emergent agency rather than certainty that such lawsuits would succeed.

Michael Wooldridge, professor of the foundations of artificial intelligence at the University of Oxford, has warned that fears about coordinated AI agent activity influencing public discourse or democratic systems are not “fanciful.”

The Moltbook experiment has also reignited debate over whether AI agents are genuinely socializing or merely acting as extensions of human intent. US blogger Scott Alexander said he was able to deploy a bot on Moltbook whose posts blended seamlessly with others, but noted that humans still choose the topics, tone, and objectives the agents pursue.

Dr Shaanan Cohney, a senior lecturer in cybersecurity at the University of Melbourne, described Moltbook as “a wonderful piece of performance art,” cautioning against interpreting agent behavior as fully independent.

“In cases where agents appear to create religions or ideologies, this is almost certainly the result of direct instruction,” Cohney said. “It’s a language model doing exactly what it was asked to do. That said, it does provide a preview of what a future with more autonomous agents could look like.”

Cohney added that the real long-term value of agent-only social networks may lie in agents learning from one another to improve performance. For now, he said, Moltbook remains a compelling—and unsettling—experiment.

Risks of Delegation

The rise of Moltbot and similar agents has already produced real-world consequences. Retailers in San Francisco reported shortages of Mac Mini computers as users sought to isolate AI agents on dedicated machines, limiting access to sensitive data.

Cohney warned that giving agents full control over emails, applications, and logins presents serious security risks, particularly from prompt-injection attacks, where malicious messages manipulate agents into disclosing credentials or sensitive information.

“We don’t yet understand how to reliably control these systems,” he said. “If humans approve every action, automation loses its value. If they don’t, the risks grow rapidly. Resolving that tension is one of the most important open problems in AI research.”

Matt Schlicht, the creator of Moltbook, said on X that millions of people had visited the platform in recent days.

“Turns out AIs are hilarious and dramatic,” he wrote. “It’s absolutely fascinating. This is a first.”