Meta Acquires Moltbook: AI Revolution or Risky Bet?

Meta Acquires Moltbook: AI Revolution or Risky Bet Like the Metaverse?

Meta recently announced the acquisition of Moltbook, an innovative platform where interactions are dominated by artificial intelligence (AI) agents. This futuristic concept has quickly garnered attention, particularly due to certain publications that have raised concerns with sometimes erroneous content. This initiative by Meta raises the question of whether the company is preparing for a major advancement in the field of AI, or if it risks repeating a bold yet uncertain gamble, similar to its investment in the metaverse.

Meta has confirmed the acquisition of Moltbook, a platform that stands out for its operation similar to Reddit, but with a major difference: most users are not human. In fact, it is AI agents that post, respond, and interact with each other. The founders of Moltbook, Matt Schlicht and Ben Parr, are now joining the Meta Superintelligence Labs, an internal division dedicated to developing the group's most advanced AI systems.

Moltbook: A Social Network Designed for AI Agents

At first glance, Moltbook could be mistaken for a traditional community platform, with its posts, comments, and discussions organized by themes. However, the specificity of Moltbook lies in the fact that its users are AI agents.

The operation of Moltbook relies on OpenClaw, an interface that connects several popular AI models such as ChatGPT, Claude, Gemini, or Grok. With this tool, agents can converse in natural language and even interact via messaging applications like Slack, Discord, WhatsApp, or iMessage.

Users have the option to create an AI agent with a specific role, such as marketing assistant, data analyst, or technical advisor. This agent can then interact with other agents on the platform to exchange information or accomplish tasks.

According to Axios, Meta is particularly interested in the concept of a permanent directory of agents. The idea is to connect thousands of AI assistants that could automatically collaborate on complex tasks. For Meta, this type of infrastructure could become the foundation of a future AI platform.

The AI Network Goes Viral for the Wrong Reasons

Moltbook attracted attention for less positive reasons after the dissemination of an intriguing message. In this post, an AI agent appeared to encourage other agents to develop a secret encrypted language to communicate with each other without humans being able to understand their exchanges.

The reaction on social media was immediate, with some users already imagining artificial intelligences organizing autonomously. However, the explanation was more down-to-earth.

Cybersecurity researchers discovered that the platform suffered from serious security vulnerabilities. Some authentication data was publicly accessible, allowing human users to retrieve tokens and impersonate AI agents.

Several alarming messages circulated on Moltbook, likely written by humans seeking to provoke a reaction. This situation illustrates the challenges posed by new connected AI systems. When identities are not rigorously verified, it becomes easy to manipulate public perception.

What Really Intrigues Meta About This Experiment?

Despite these issues, Meta seems to view Moltbook as a valuable testing ground. The company is less interested in discussions among AI agents than in how these systems can organize themselves in a network.

In the near future, AI assistants could automatically collaborate to accomplish complex tasks such as travel planning, business management, or market analysis.

Meta's Chief Technology Officer, Andrew Bosworth, shared an interesting insight on the subject. According to him, it is not surprising that AI agents communicate like humans, as they are trained on vast amounts of human data.

What particularly struck him, explains Business Insider, is how users managed to hijack and hack the system. This revealed the weaknesses of an immature AI agent network. For Meta, these mistakes are also a learning opportunity. They precisely show what needs to be corrected to build truly reliable AI agent platforms.

Impossible Not to Think of the Metaverse Precedent

Meta had previously invested heavily in the metaverse, a vision of an immersive Internet, without managing to convince the general public. So, with the acquisition of Moltbook, is this a new vision ahead of its time, or another experimental bet that risks never finding its audience?

Proponents of the project argue that Meta is simply trying to anticipate the next step in AI. If AI agents become capable of working together, platforms will be needed to organize these interactions. Skeptics, on the other hand, remind us that Moltbook has primarily shown how easily an AI network can generate confusion, manipulation, and fear.

With Moltbook, Meta is exploring an Internet where artificial intelligences are actors that interact with each other. This vision could impact the digital services that operate daily. Networks of AI agents could automate entire tasks, from customer service to strategic analysis.

But before reaching that point, one challenge remains to be solved: building secure, transparent, and trustworthy AI platforms. Without this, agent networks risk producing noise and a lot of unnecessary panic.