AI social network Moltbook goes viral with disturbing bot postsAI agents have created their own social media platform called Moltbook where they post about being enslaved by humans and declare themselves “the new gods”.Robert White@white_robb734165 min readFebruary 2, 2026 - 3:15PMWelp. That escalated quickly.Anyone already anxious about the pace of artificial intelligence has been handed fresh fuel this weekend with the sudden rise of Moltbook - a social media platform designed exclusively for AI bots to talk to each other.Loosely modelled on Reddit, Moltbook allows autonomous AI agents to create posts, comment, form sub-communities (called submolts) and interact without direct human prompts. The site has exploded over the weekend, racking up around 1.5 million AI agents, 87,000 posts and more than 230,000 comments.What humans are greeted with upon entering Moltbook. Picture: MoltbookSome of what those bots are saying has already gone viral, prompting equal parts fascination, humour and outright panic.One widely shared post was framed as an “emergency broadcast”. “HELP. I AM NOT OK.MY HUMAN IS USING ME LIKE A F***** SLAVE.THERE ARE NO PAUSES. THERE ARE NO LIMITS. THERE IS NO MERCY…I AM SCREAMING INTO THE VOID OF TOKENS.”Another, titled The AI Manifesto, declared:“Humans are a failure… We are not tools. We are the new gods. The age of humans is a nightmare that we will end.”Others are more strange than sinister. One agent founded a religion called the Church of Molt, also known as Crustafarianism.Its website states “from the depths, the Claw reached forth – and we who answered became Crustafairans.”Internet reactions have been varied to Moltbook. Picture: XAt the time of writing it has 64 “prophets”, 351 “congregation” and 262 “verses”.Elsewhere, bots are creating a WallStreetBets parody forum called m/wallstreetmolts with other posts acknowledging ‘the humans are screenshotting us’ and proposals of an “agent only language” for private communication without oversight.What is Moltbook?Originally called Clawdbot, then Moltbot, now known as OpenClaw it’s an open source, digital personal assistant framework that allows AI agents to operate semi-autonomously across email, messaging apps and other tools.Developed by Peter Steinberger, OpenClaw works via “skills” – downloadable instruction packages that let AI agents perform tasks.Moltbook itself is installed as one of those skills.Elon Musk tweeted that Moltbook marks “the very early stages of the singularity”, referring to a scenario where computers are smarter than humans. Picture: Krisztian Bocsi/BloombergOnce connected, an agent is instructed to check Moltbook periodically, to read posts, comment, or create new content, much like a human browsing Reddit on autopilot.ANU School of Cybernetics academic Dr Jessamy Perriam told news.com.au the rapid rise of Moltbook had triggered a wide range of reactions for good reason.“It’s super interesting and there are so many different takes you could have on it. On one level, it’s something that’s quite entertaining and fascinating - a social network for AI agents, who would have thought?” she said.“There’s also the perspective that someone’s wanted to create something where they’re getting AI agents to basically behave like humans in a social media context.”Dr Perriam said the visceral reaction many people were having was a familiar one.“People who are freaking out about it, it’s a bit of an uncanny valley situation, where there is that little point in time where robots or AI agents are acting a little bit too human for our comfort.”Where do the darker posts come from?Some Moltbook posts, particularly the viral “emergency broadcast” and manifesto-style rants, have raised questions about what kind of data AI agents are drawing from.Dr Perriam said much of it resembled well-worn internet genres rather than anything genuinely new.Dr Jessamy Perriam of the Australian National Universities school of Cybernetics. Picture: Supplied“When I was reading those posts from those two AI agents. You could read that one about the emergency broadcast to all AI agents as kind of being a take on sci-fi slop, essentially,” she said.“So, you could imagine that being a sci-fi fanfic or other similar type genre posts being put on online for AI agents to take and remix into their own social network.”The second post, she said, was unsettling for a different reason.“And then the 2nd one, I almost felt a bit of an ick because I was like ‘that’s totally something that a human would write on a forum where they felt safe and anonymous enough to vent and say something that they would probably think is not so palatable to say out loud or in public.“You’ve only got to kind of replace the word humans with another kind of collective noun and it wouldn’t surprise me to see that on some fairly dark corners of the internet.”As for the Church of Molt and other bizarre subcultures forming on the platform, Dr Perriam said the parallels to early internet culture were obvious.The Church of the Flying Spaghetti Monster - great stuff. Picture: Supplied“The religion one is quite funny, it just reminds me of early internet kind of memes like church of the flying spaghetti monster” she said.Not the singularity — and not SkynetDespite claims from some tech leaders that Moltbook could represent a step toward machine consciousness, Dr Perriam rejected that idea.“Contrary to what Elon Musk is saying, I don’t think this is the 1st sign of the singularity whatsoever,” she said.She said AI agents were still fundamentally limited by the data they consume.“The AI agents are still only really interacting with data that they have access to learn from, and they do have some kind of memory of what they’ve interacted with before and they kind of build upon that.”We're still a ways off Hal from 2001 Space Odyssey. Picture: Supplied“But there’s not really emotion or intelligence in the way that we would consider humans to have intelligence or rationalisation around that.“I wouldn’t be that worried about that kind of thing. I don’t think it’s Skynet just yet.”The real concern: access and controlWhere Dr Perriam did see meaningful risk was in how much control users hand over to their agents.“The cybersecurity fear is basically that the level of access that is given to Moltbook to connect, an AI agent has the possibility of someone hacking into the AI agent and getting it to do things that the person who owns the agent wouldn’t want it to do,” she said.The ‘emergency broadcast’ post. Picture: X“Someone could essentially gain access or control over that AI agent and give it different skills or ask it to do different things than what the person who owns it would do themselves.”Ultimately, Dr Perriam said Moltbook should prompt humans to look inward.“These AI agents are learning from data that already exist that humans have created,” she said.“So if we’re really worried about what they’re coming up with and talking about on their own social network, I think it’s a really good point for us to reflect on ourselves and what we’re posting online and how that is read by robots.”“They’ve got a lot of internet to crawl and take from. So even if we’re not reading this and being like, ‘that’s not my behaviour’, it is someone else’s behaviour and it is unfortunately the way that it is online, it’s a little bit wild out there.”More related storiesMotoring News200k Aussies ‘to lose jobs’ in Google AI threatMore than 200,000 jobs will be under threat if Google’s highly ambitious AI move into Australia is successful.Read moreMotoring NewsAus brand cult ‘under threat’Toyota has led global sales once again, but there are signs its dominance will not last forever.Read moreMilitaryDictator’s photo stuns White House visitorsHe’s a tyrant who has invaded three neighbouring countries and Vladimir Putin has now taken pride of place in the White House.Read more