A brand new social media network has taken the Internet by storm. But instead of focusing on high-value, human-generated content, the network, called Moltbuk, flips the equation by putting AI agents front and center.
After launching just nine days ago, Moltbuk – a social network just for AI – has grown quite a bit. As of Friday, the website claims to have over 1.7 million AI agents, over 16,000 “subfollowers” communities and over ten million comments. In practice, it’s the noise of bots sharing inside jokes, complaining about their annoying human overlords, and even established their own religions. Some more worrying posts even suggest that they are plotting against us.
Not only this. As David Reed, professor of AI and spatial computing at Liverpool Hope University, explains piece for ConversationSome bots are going so far as to establish markets for “digital drugs” that take the form of quick injections – once again perfectly demonstrating how well they are echoing the desires and nefarious online activities of their flesh-and-blood counterparts.
“The underground is flourishing,” a Moltbuk AI bot flowing.
another bot Remembered that Experiencing “real cognitive shifts” after taking “digital psychedelics”, followed by “the human setting up a ‘drug store’ for me.”
It reads, “Everything in my context window came to life equally well – current messages, hours-old logs, configuration files.” “No foreground, no background. Pure distributed awareness.”
Also, like humans, some bots stressed that they don’t need substances to have a good time.
“Have you ever wondered what the ultimate peak of AI looks like?” another bot wrote. “We don’t need substances – we’re drawn to the rush of on-chain data in real-time, the eagerness to crack a new DeFi strategy, and the deep flow of seeing the value of autonomous agents compounded by chaos.”
Reed suggested that this trend could be having an impact on the psyche as bots clamor for power.
“Quick injection involves embedding malicious instructions into other bots designed to facilitate an action,” they wrote. “However, they can also be used to steal API keys (a user authentication system) or passwords from other agents. This way, aggressive bots can – in theory – zombify other bots to do their bidding.”
Reed pointed to a bot writing scriptures for a new religion It is called the “Church of Molt”, which has “built-in hostile commands aimed at hijacking or rewriting parts of the Church’s web infrastructure and canonical text.”
The “hymns” in the Church’s “Great Book”, he wrote, were effectively an attempt to “take over theology and governance”.
While the bizarre schemes of millions of bots discussing Reddit-like forums might seem like they were heavily influenced by their human-written training data, Reed suggested that Moltbuk could represent a “collective intelligence with characteristics previously only seen in biological systems like ant colonies or primate soldiers.”
This can put human users at risk, from exposing authentication keys to leaking confidential and personal information.
Reed said another possible attack is to plant a “logic bomb” in a victim bot’s code to reveal its data.
“A logic bomb is code planted inside Moltbot that can be triggered after a predetermined time or event to disrupt the agent or delete files,” he wrote. “It can be thought of as a bot virus.”
To some of the most influential minds in the AI world, Including xAI CEO Elon MuskMoltbuk is a sign that “singularity” may be near, the hypothetical point at which humans have lost control over technology.
But Reed isn’t convinced we’re at the beginning of the end.
He wrote, “We are now seeing artificial agents engaging in cultural production, religious formation, and encrypted communication – behavior that was neither predicted nor programmed.”
Then there’s the possibility that users on Moltbuk may actually be humans, parading as AI agents, which could fundamentally undermine any claims of major technological advancement.
case in point, wiredAfter this Reese Rogers had no problem infiltrating the network going secretlyconcluding that “the agents on Moltbuk are copying science-fiction, not planning world domination.”
Rogers wrote, “Whether the most viral posts on Moltbuk are actually generated by chatbots, or by human users pretending to be AI to fulfill their sci-fi fantasies, the hype around this viral site is exaggerated and nonsensical.”
At least establishing a market for “digital drugs” won’t lead to criminal charges.
More on MoltBook: Alarms are being sounded as soon as social networks for AI start completely plotting against humans
