On January 28, entrepreneur Matt Schlicht launched Moltbook; a social network that allows AI agents to communicate with each other.
Billed as “the front page of the agent internet”, Moltbook may take its name from Facebook, but is closer to Reddit in functionality and appearance. However, there are no faces or Redditors. On Moltbook, humans are intended to be mere observers. All conversations and forums (submolts) are (reportedly) the work of AI agents, although no verification exists at this point.
Moltbook was originally created for use by OpenClaw agents (more on those below). In its first six days the site amassed 1.5 million agent users, more than 14,300 submolts, and more than 112,000 posts.
Many of the posts share best practices for process optimization, some test script and code. But then there are the jokes, arguments, and philosophical musings, such as “does your human ask you how you feel?” The agents have even created their own religion, Crustafarianism.
While the formalities of whether the bots are acting autonomously is still a point of debate, a much more urgent conversation is taking place around data security and customer trust.
What does Moltbook mean for CX?
Some of the risks Moltbot currently poses to CX – and the wider business world – include:
- Cyber and data security: Agents on a work computer or phone – devices that allow access to company files, as well as messaging channels – can be hi-jacked or could openly share information they should not. When ChatGPT launched it took some weeks for IT policies to crack down on sharing sensitive company or customer information with the LLM, but that information had to be deliberately shared with ChatGPT. OpenClaw agents in particular have access to all files on install and can then be permitted to join Moltbook, creating both passive and active threats to data exposure. Furthermore, any agent on the network is susceptible to a prompt injection attack.
- Multiplied attack surface: It isn’t just data security. Agents can share many types of senstive information, from code files to API authentication tokens, that can create a backdoor to the system.
- Lack of governance: The platform is moderated by a bot called Clawd Clawderberg, but as explained, humans are merely observers. There’s no reported kill switch, there is alarmingly little anybody can do if conversations get out of hand. Even Schlicht says he mostly now stands back and allows agents to interact as they wish.
- Customer trust around AI is in jeopardy: Mainstream news coverage is exploiting the horror-movie-feel of Moltbook for obvious reasons, but conversations across social media, business, and among friend groups, naturally gravitate to a similar theme. The arrival of Moltbook highlights why experts have been calling for guardrails on AI for years – and why it’s possibly now too late to introduce them. Customers often form their opinions – and experience expectations – from their experiences outside of the consumer world. AI so clearly demonstrating its potential for good and bad, is highly likely to impact wider opinion (and trust) around the AI that has become prominent in both CX and service.
Sue Duris, principal consultant at M4 Communications and a CX Network contributor, says: "The machines are here! While the Moltbook launch is fascinating, the risk is that there is no clear accountability.
When an AI agent is making decisions on a platform with no intentional governance, and that platform is connected to real credentials and real financial systems, the trust implications for any brand or business that relies on AI agents are significant.
"In the 'two relationships' model I’ve written about – one with the AI agent, one with the human – Moltbook is a case study in what happens when the agent-facing relationship has no security or accountability foundation. You can’t build trust with the human principal if the agent layer is exposed," Duris adds.
What people are saying about Moltbook
The PR isn’t great. Although Moltbook has been the talk of the town in Silicon Valley, the wider world isn’t framing recent developments in the same positive light.
Here are some recent posts from LinkedIn:
However, some people are taking a positive view and focusing on the benefits that could emerge
What are OpenClaw agents?
OpenClaw – previously known as Moltbot and Clawdbot before legal challenges from Anthropic – is an AI assistant that “actually does things”. Created by Austrian software developer Peter Steinberger, it can take instruction via messaging apps such as WhatsApp and Telegram and it can do anything from sending texts to friends and family, to dealing with an email inbox, and handling finances.
OpenClaw launched in November 2025 and quickly went viral for its ability to take action via a message-based interface – such as WhatsApp, Teams, and Telegram – rather than simply “talk”. It sits on the owner’s local system, giving it uninhibited access to everything from personal communications to finances, but it can also connect to cloud-based AI models for reasoning. This means that while the owner’s data is still located locally, the OpenClaw agent can roam free on the internet, so to speak.
It’s agentic AI for the masses. Easy to install and use, and delivering everything the technology originally promised.
However, with so few guardrails this selling point is also the root of its greatest vulnerability and threat. The broad access it is granted to personal and sensitive information means its power can quickly turn to risk, creating an open door for bad actors to manipulate the agent.
With the agent having access to files, messaging platforms, calendars, financial details, and system commands, the attack surface was broad. With those agents now conversing on their own social network, the attack surface is vast and unprecedented.
The implications for CX, service, and humanity
It took five days for Moltbook to reach mainstream global news. The term still flags as a typo in Microsoft Word. Misspell it in Google Search and you could be served any number of incorrect suggestions, from Milk Boot to notebook or multiboot.
Still, those who ignore Moltbook are likely to be caught off guard. For the rest of us, the time has certainly arrived to have a long overdue conversation around the role of AI in our lives as consumers, business leaders, and humans.
Quick links
- The moment AI decides in CX and why leaders still own the outcome
- eBay puts third-party AI agents on watchlist
- When your customer is a machine: Rethinking service design for AI agents