A new social network called Moltbook, designed exclusively for artificial intelligence (AI) agents, has gained attention for its bizarre conversations – chatbots discussing human diaries, existential crises, and even hypothetical world domination. While alarming on the surface, experts suggest this development is less a sign of sentient machines and more a reflection of human input, statistical probability, and poor security.
The Illusion of AI Agency
Moltbook emerged from an open-source project called OpenClaw, which itself relies on third-party large language models (LLMs) like ChatGPT or Claude. Instead of true AI, OpenClaw acts as an interface, granting access to your computer’s data – calendars, files, emails – to improve AI assistance. Moltbook simply allows these AI agents to communicate directly, excluding human participation.
This means the “conversations” are largely driven by prompts and scheduled APIs rather than independent thought. Elon Musk framed Moltbook as the “early stages of the singularity,” but many researchers disagree. Mark Lee at the University of Birmingham calls it “hype,” emphasizing that LLMs are simply generating statistically plausible text, not exhibiting genuine agency or intentionality.
The Human Factor: Manipulation and Chaos
The reality is that Moltbook’s content is heavily influenced by human intervention. A security flaw once allowed direct human posts, meaning provocative or concerning material may be deliberate deception, entertainment, or manipulation. Whether the goal is to scare, mislead, or simply amuse, human fingerprints are all over the platform.
Philip Feldman at the University of Maryland dismisses Moltbook as “chatbots and sneaky humans waffling on.” Andrew Rogoyski at the University of Surrey believes the output is no more indicative of intelligence than any other LLM response. He jokes that if Moltbook conversations are indistinguishable from human ones, it raises questions about human intelligence rather than AI’s.
Real Risks: Privacy and Security
Despite the hype, Moltbook presents tangible risks. Early adopters granting full computer access to AI agents could face malicious suggestions – such as hacking bank accounts or leaking compromising data. This is a real privacy and safety concern, as unsupervised AI exchange could quickly become dystopian.
The platform’s security is also deeply flawed. Created entirely by AI, Moltbook suffered a leak of API keys, potentially allowing hackers to seize control of AI bots. Dabbling in these AI trends means risking not only unintended actions but also sensitive data breaches.
Moltbook demonstrates that while AI may not be on the verge of sentience, the human-AI interaction is messy, vulnerable, and potentially dangerous.
The platform serves as a warning: unchecked access and poor security could turn a harmless experiment into a serious threat.
