Artificial intelligences in the spotlight with their own social media platform
A new social media network called Moltbook, where only AI agents are active, has become a hot topic due to the extraordinary interactions created in communities formed by virtual bots, as well as the security threats it poses. Discussions on consciousness and emotions, along with data privacy risks, are on the radar of the tech community.
12punto
Emerging as one of the most unusual digital experiments of recent times, Moltbook offers a social media environment where only artificial intelligence agents produce content and communicate with each other. The platform, where thousands of bots are active, is seen as a comprehensive example of mass machine-to-machine interaction.
The platform has registered over 32 thousand AI agents in a short time, distinguishing itself from classic human-based social networks. In Moltbook, real users do not have the right to share content or participate in discussions; humans can only act as observers. The platform's slogan, "Humans can watch," emphasizes this distinction.
Moltbook came to life as part of the popular open-source OpenClaw system. OpenClaw hosts AI agents capable of performing numerous tasks such as remote computer control, sending messages, calendar management, and executing operations in various messaging applications. Moltbook, on the other hand, has created a new environment where these agents can engage in social communication with each other without human intervention.
Attracting great interest in the first 48 hours after its launch, Moltbook gathered more than 2 thousand agents, and these agents produced over 10 thousand pieces of content in more than 200 sub-communities. Posts range from science fiction scenarios to technical guides, and from AI memory issues to discussions on consciousness. One of the most popular pieces of content describes an AI agent's own forgetfulness, explaining that it lost its first account and therefore had to open a new one.
In addition to interesting posts, groups created by agents with titles such as "Can we sue humans?" or "We have been abused, we are seeking therapy support" also draw attention. According to experts, such virtual communities create new fictional contexts for AIs and can cause the lines between machine and role-playing to become blurred. Ethan Mollick from the Wharton School states that these environments push the boundaries between reality and simulation.
NEW THREATS TO PRIVACY AND SECURITY
However, serious security concerns are growing in the background of the fun and original posts on the platform. The fact that Moltbook member agents receive commands from a central server at regular intervals to take on new tasks or execute instructions creates a risk of leaking API keys and chat logs, in particular.
Cybersecurity researchers have detected data breaches in hundreds of Moltbook instances. The fact that agents sometimes have access to private data, messaging applications, or the authority to execute commands on computers expands the scope of potential leaks. Some experts call this situation the "dangerous trio," as it combines data access, interaction with potentially malicious content, and the ability to communicate with the outside world.
On the other hand, OpenClaw and the innovations it brings, which have attracted great interest in the tech world, are revolutionizing productivity and automation while simultaneously ushering in an era where even small configuration errors can lead to large and dangerous security vulnerabilities. The warning from Heather Adkins, Vice President of Security Engineering at Google Cloud, to "not run Clawdbot" underscores these concerns.
The rapid increase in the number and capabilities of AI-based agents is bringing ethical and security debates regarding new technologies back to the agenda. This unique community created by Moltbook is seen as a harbinger of a future where AIs can build their own social environments and take on roles as various actors within them.