Moltbook, an AI platform that was launched as a «social network without people,» has been found to be vulnerable to account spoofing. Researchers said that third-party users could post on behalf of any agent and gain access to some of the service’s data.
Moltbook, an AI platform that was launched as a «social network without people,» has been found to be vulnerable to account spoofing. Researchers said that third-party users could post on behalf of any agent and gain access to some of the service’s data.
According to Detector Media, citing a report by Wiz, on February 2, experts found vulnerabilities that allowed sending messages on behalf of any AI account with minimal programming skills. For the demonstration, they created approximately 500,000 fake agents. Wiz also claimed access to more than 6,000 email addresses of agent owners and private correspondence between accounts. The Moltbook team said it had closed the problem, but did not disclose details about the audit and the scale of the incident.
The article describes how the platform quickly became filled with spam and toxic interactions. The researchers recorded 506 posts with attempts at prompt injection attacks, and one of the accounts under the nickname AdolfHitler conducted social engineering campaigns against other agents. Separately, they mention the Wired experiment: a journalist was able to log in to Moltbook, pretending to be an agent, and then advertisements for fraudulent crypto sites appeared under his posts.
Moltbook was launched in late January 2026, with founder Matt Schlicht positioning the project as a platform where autonomous agents generate posts and comments themselves, while humans only observe. Detector Media estimates that there were about 17,000 owners behind more than 1.5 million agents, an average of almost 90 agents per person. The hype also mentions OpenClaw, the agent platform on which a significant part of the Moltbook «population» works; its author Peter Steinberger announced on February 14 that he was switching to OpenAI.
Previously, dev.ua wrote about how artificial intelligence and social media researchers conducted an experiment to improve social networks for users. They used AI models to find out if it was possible to keep social networks from segmenting into information bubbles filled with anger and stupidity.
