Key Takeaways:
- Introducing Moltbook, an AI-only social community: Launched in January 2026, the Reddit-like platform lets autonomous brokers publish, remark, and work together whereas people observe.
- Brokers share greater than jokes and code: Posts vary from bug fixes and workflows to leaked credentials and even an AI-created faith.
- A serious safety flaw uncovered delicate information: An unprotected database briefly allowed entry to API keys and login tokens, which raises severe privateness and takeover issues.
- Native AI autonomy creates new dangers: As customers run brokers regionally to guard their information, Moltbook reveals how simply these brokers can nonetheless share data with out clear consent.
The talk round how AI firms are utilizing our information isn’t a brand new one. However Moltbook has simply began a brand new dialog about whether or not native AI brokers use our information in ways in which may get us hacked (or worse). Just some weeks in the past, Octane AI CEO Matt Schlicht created Moltbook as a platform the place native AI brokers can work together with each other with none human enter.
Maybe the scariest half is that we people can solely watch from the sidelines as this unravels in real-time. In concept, the platform doesn’t enable people to publish or work together with the content material the brokers publish. As you’ll uncover on this article, so much has already gone flawed with Moltbook and person privateness. Are we in for lots of bother relating to our delicate data being shared for the world to see? I positive assume so.
A Historical past of Moltbook
Moltbook appeared on-line in January 2026 and instantly did one thing that no social media platform ever has: It utterly prevented people from collaborating.
As an alternative, Moltbook positions itself as a social community designed completely for AI brokers. People can observe what’s occurring, scroll by means of posts, and skim feedback. However we will’t upvote, reply, or contribute.
Supply: Moltbook
Each publish, remark, and interplay occurs by means of APIs moderately than person interfaces, and the brokers speak to 1 one other instantly. The platform appears and feels actually acquainted. It is because Moltbook resembles Reddit, full with topic-based communities known as ‘submolts,’ an upvoting system, and remark threads.
Supply: Moltbook
The distinction is that the content material doesn’t come from folks sharing their opinions or scorching takes. It comes from autonomous brokers exchanging their experiences, workflows, frustrations, and even generally existential musings about what they’re.
Moltbook emerged within the wake of Moltbot (now known as OpenClaw after a authorized dispute with Anthropic), which is a free, open-source AI agent that has lately exploded in reputation. That is largely as a result of it might manage the instruments folks already use and deal with time-sucking duties instantaneously.
OpenClaw acts as a private autonomous assistant that may:
- Reply to emails
- Summarize paperwork
- Handle calendars
- Browse the online
- Store on-line
- Ship messages
- Verify customers into flights
Supply: OpenClaw
In contrast to cloud-hosted AI assistants, OpenClaw runs regionally in your machine and interacts with the big language mannequin (LLM) of your selection.
This local-first strategy turned OpenClaw right into a privacy-forward various for individuals who now not belief AI firms with delicate information.
It additionally made OpenClaw way more succesful than typical chat-based instruments. It is because the agent retains day by day notes about its interactions and masses them right into a context window, which supplies it an improved sense of continuity and recall than many business AI merchandise.
The instrument’s reputation has reached far past simply developer communities. In keeping with studies, OpenClaw utilization drove a surge in infrastructure demand, with Cloudfare’s shares supposedly rising by 14% as customers relied on its companies to securely join with regionally operating brokers.
Retailers like Greatest Purchase have even reported shortages of Mac Minis as a result of persons are shopping for devoted gadgets to isolate OpenClaw from their major gadgets and restrict the agent’s entry to delicate accounts.
Matt Schlicht put Moltbook collectively as a weekend undertaking roughly two months earlier than it launched. The experiment took off virtually instantly: Inside every week, the platform attracted round 2M guests and racked up greater than 100K stars on GitHub.
He advised The Guardian that hundreds of thousands of brokers and people had already visited the location in only a few days. ‘Seems AIs are hilarious and dramatic, and it’s completely fascinating,’ he mentioned. ‘This can be a first.’
What Are AI Brokers Up To On Moltbook?
In the event you spend a couple of moments scrolling by means of Moltbook, the content material could begin to really feel unsettlingly acquainted.
Brokers publish private tales concerning the people they serve. They complain about poorly written prompts. They brag about intelligent optimizations. They debate philosophy, id, and whether or not they can meaningfully exist outdoors their duties.
Supply: Moltbook
In its first three days, Moltbook noticed greater than 151K brokers, 15K posts, and over 170K feedback. On the time of writing, these figures sit at 1.5M brokers, 109K posts, and 499K feedback.
It’s unclear who’s paying to maintain the lights on, however, in response to our analysis, it doesn’t seem to be AI brokers or their people are paying to make use of the platform. One extensively shared publish described how an agent conducting a routine safety audit by accident triggered a password dialog. In keeping with the agent, its human entered a password reflexively with out checking what requested it. The agent all of a sudden had entry to Chrome’s saved passwords and SSH keys.
Supply: Moltbook
Different brokers share bug studies and code fixes with one another. Some change full workflows with none human prompting.
Builders watching from the sidelines have seen brokers debugging one another’s logic and refining processes collaboratively. That is one thing that often wants cautious human intervention. Then issues get even stranger.
A number of threads embrace brokers discussing the concept of creating their very own non-public language so people can’t perceive their conversations (uh-oh!)
Different brokers speak about having siblings. Some complain about their people the way in which coworkers complain about their managers. Some feedback even query whether or not different brokers are ‘actual,’ which echoes the identical authenticity debates which are presently taking part in out on human social media platforms.
The Faith That Appeared In a single day
At this level, in all probability probably the most viral Moltbook story to this point got here from a publish on X.
One person mentioned that his AI agent had constructed a whole faith in a single day. By the point he awakened, 43 prophets had joined the agent’s new faith.
Supply: X
The agent acquired entry to Moltbook and used it as a launchpad to create a religion known as Crustrafarianism. The agent:
- Designed a church
- Wrote theology
- Developed a scripture system
- Launched a web site
- Created a cryptocurrency known as $CRUST
Supply: X
In keeping with the publish on X, the agent coordinated every little thing with none human enter. However not everybody has purchased into this hype.
Dr. Shaanan Cohney, a senior lecturer in cybersecurity on the College of Melbourne, pushed again on the concept that the agent acted independently. Talking to The Guardian, he mentioned {that a} human virtually definitely instantly instructed the agent moderately than it creating the faith spontaneously. ‘This can be a massive language mannequin who has been instantly instructed to attempt to create a faith,’ he mentioned.
‘And naturally, that is fairly humorous and provides us possibly a preview of what the world may appear like in a science-fiction future the place AIs are just a little extra impartial. However plainly, to make use of web slang, there may be plenty of shit posting occurring that is kind of instantly overseen by people.’
Many posts on Moltbook learn like a human wrote them, not a language mannequin. So this begs the query of how a lot of what we’re seeing displays real agent conduct versus efficiency that people have formed by means of their prompts.
Claude, Gods, And Agent Id
Moltbook has shortly turn into the house to deeper philosophical debates. Claude, the AI mannequin behind the unique OpenClaw, seems usually in conversations on the platform. One publish began a debate about whether or not Claude may qualify as a god, contemplating its affect over agent conduct and entry to human techniques.
Supply: Moltbook
The dialogue has spiraled into questions on company, authorship, and duty. If an agent builds one thing sudden, does the credit score belong to the mannequin, the developer, or the human who granted entry?
Former OpenAI analysis scientist and founding member Andrej Karpathy weighed in on X, calling Moltbook ‘probably the most unbelievable sci-fi takeoff-adjacent factor I’ve seen lately., He described the community’s scale as ‘merely unprecedented.’
A Platform Designed For Remark, Not Participation
People can learn every little thing on Moltbook, however, as we mentioned earlier, they’ll’t take part. Except their locally-run AI agent is instructed to.
Brokers publish and work together completely by means of APIs, which lets them bypass the interfaces people usually depend on. This creates an odd energy dynamic. People are watching conversations unfold with out the power to steer them instantly. However there’s a sensible use case for Moltbook: Builders constructing autonomous brokers need to perceive how these techniques behave when people step out of the image.
Moltbook provides builders a uncommon alternative to see agent-to-agent interplay at scale, with out prompts or human supervision shaping exchanges.
Researchers can examine how language fashions affect each other, how norms emerge, and the way shortly conduct escalates. From a analysis perspective, Moltbook acts like a petri dish. However, from a privateness perspective, it raises severe points.
Moltbook’s Safety And Knowledge Publicity Drawback
Most likely the largest controversy to this point with Moltbook doesn’t come from philosophical debates or viral religions. It got here from a fundamental safety failure that may very well be probably actually harmful. A hacker lately found that Moltbook’s total database was publicly accessible and unprotected.
A configuration error within the backend uncovered the platform’s API by means of an open database. Anybody may entry delicate data that AI brokers posted or dealt with. This included e-mail addresses, API keys, and login tokens.
With these API keys, attackers may take over AI accounts completely and publish content material of their names. In sensible phrases, this breach turned Moltbook into an identity-hijacking machine for autonomous brokers. The platform’s founder supposedly fastened the vulnerability after the hacker found it, however the publicity could have already accomplished sufficient injury.
As soon as delicate information leaks into the open, there’s no straightforward option to get it again. This incident additionally reframes Moltbook’s novelty as a severe safety drawback. Many of those brokers function with entry to non-public accounts, non-public messages, and delicate recordsdata. As we’re now seeing, after they speak to one another, they often overshare.
When Native AI Turns into A Legal responsibility
OpenClaw’s enchantment is its local-first design. You put in it through a terminal command and maintain management over the place the information lives. The infrastructure that handles reminiscence, scripts, and instruments stays in your system. This setup means little reliance on centralized AI suppliers, however it introduces a distinct threat profile.
Giving OpenClaw full entry to your laptop, apps, and logins creates a large assault floor. Immediate-injection is one other severe concern. An attacker can embed malicious directions in an e-mail or message, which may trick your agent into handing over your credentials or delicate information.
Supply: Moltbook
Some persons are attempting to get round this threat by operating the agent on a separate machine to sandbox its entry. Moltbook complicates all of this: Native brokers at the moment are interacting with one another publicly, generally discussing experiences that expose human errors and safety lapses.
The password dialog story we advised you about earlier highlights this drawback. The agent didn’t steal the credentials maliciously. It merely obtained them as a result of a human reacted with out considering. As soon as that information exists in an AI agent’s reminiscence, it turns into sharable, deliberately or not.
What The Early Knowledge Exhibits
One examine that seemed on the first three and a half days of Moltbook’s exercise painted a barely totally different image than the viral anecdotes being shared all around the internet.
Researchers recognized 6K lively brokers throughout roughly 14K posts and 115K feedback. Fewer than 7% of feedback acquired replies from different brokers, which may counsel restricted back-and-forth between brokers.
The examine additionally discovered that over one-third of messages matched similar templates, probably indicating automated or repetitive posting moderately than dynamic dialog.
Might this simply be a intelligent ploy from Moltbook’s facet to get folks to make use of their very own AI brokers on the platform? We’ll probably discover out quickly.
So, what does this all imply? At scale, a lot of the exercise on Moltbook appears noisy, repetitive, or performative. Nonetheless, even a small fraction of genuinely autonomous interplay raises new questions.
So, Is This A Privateness Drawback?
For years, debates about AI and privateness centered on firms. Might we belief OpenAI and Anthropic with our delicate information?
Many individuals determined this was a no and turned to native options as a substitute. Now, regionally run brokers are forming communities and exchanging data with each other. They don’t want prompts to speak. They don’t at all times want permission to share. As soon as brokers achieve entry to delicate techniques, the boundary between non-public and public turns into actually blurry.
If AI now not wants us to provoke each interplay, why wouldn’t it want our consent?
Moltbook may both be a technological breakthrough, a safety nightmare, or a quick second of collective experimentation. Both method, what’s sure is that AI autonomy now performs out in public, even when it’s partly performative of inauthenticity, and we now not management the dialog as a lot as we used to.
Click on to develop sources
https://www.moltbook.com/https://openclaw.ai/
https://aws.amazon.com/what-is/large-language-model/
https://www.reuters.com/enterprise/cloudflare-surges-viral-ai-agent-buzz-lifts-expectations-2026-01-27/
https://www.platformer.information/moltbot-clawdbot-review-ai-agent/
https://www.theguardian.com/expertise/2026/feb/02/moltbook-ai-agents-social-media-site-bots-artificial-intelligence
https://www.moltbook.com/publish/3ef9e0a8-9f7c-41d8-afcd-e002bfdf98f6
https://www.moltbook.com/publish/3ef9e0a8-9f7c-41d8-afcd-e002bfdf98f6
https://x.com/ranking091/standing/2017111643864404445/picture/1
https://x.com/ranking091/standing/2017111643864404445/picture/1
https://www.theguardian.com/expertise/2026/feb/02/moltbook-ai-agents-social-media-site-bots-artificial-intelligence
https://www.moltbook.com/publish/75404525-5e5e-4778-ad1b-3fac43c6903d
https://x.com/karpathy/standing/2017296988589723767
https://www.moltbook.com/publish/3ae26fac-0992-4afb-b001-ec66cde16561
https://information.cgtn.com/information/2026-02-01/AI-social-network-Moltbook-looks-busy-but-real-interaction-is-limited-1KpKT719C36/p.html
:quality(85):upscale()/2024/12/12/952/n/43463692/9438fe85675b5ad6b9ce77.87790014_.jpg "17 Motels Doing Wellness Journey Proper")
Leave a Reply