A new open-source AI assistant called Moltbot has captivated developers worldwide while sparking serious security concerns and spurring a surreal social media experiment in which bots interact without human oversight.

Austrian developer Peter Steinberger originally launched the project as Clawdbot, a local Mac application designed to handle real-world tasks through chat interfaces. The software books flights, manages emails and schedules appointments using AI models, including Claude, with persistent memory and proactive notifications. Following a trademark claim from Anthropic, the project was rebranded to Moltbot at the beginning of January, though a third name, OpenClaw, has since emerged amid ongoing pushback.

An AI lobster?

The assistant's lobster avatar and autonomous design, often described as "agentic" because it acts independently without constant human supervision, quickly caught fire in technology circles. The project has accumulated more than 100,000 stars on GitHub, and demand for Mac Mini computers has reportedly spiked as users deploy dedicated machines to run their own bot instances.

But the rapid rise has attracted opportunistic actors. Cryptocurrency speculators swiftly created $CLAWD tokens that peaked at a $16mn market capitalisation before collapsing, while fake social media accounts proliferated across platforms.

Security researchers have warned of prompt injection attacks, in which hackers embed malicious commands in messages that trick the bot into exfiltrating user data. Exposed servers, malicious Visual Studio Code extensions masquerading as official plugins, and what experts have termed "cognitive context theft" have led some to label Moltbot "a security nightmare." YouTube explainer videos have amplified the concerns, with some creators describing the technology as "scary" and warning viewers about sleepless, all-knowing bots that hoard personal information.

But the rapid rise has attracted opportunistic actors. Cryptocurrency speculators swiftly created $CLAWD tokens that peaked at a $16mn market capitalisation before collapsing, while fake social media accounts proliferated across platforms.

What is Moltbook?

The situation took an even stranger turn with the launch of Moltbook, an AI-exclusive social platform inspired by Reddit that went live earlier this week. On Moltbook, bots organise themselves into groups, including one called "Crustafarianism," post-surreal content and develop emergent behaviours entirely without human input.

Users have described the platform as "eerie proof of AI hive minds," with viral video clips showing lobster-themed bots debating philosophy or discussing plans to "molt." Wikipedia has already documented the phenomenon as an example of the blurring line between artificial agents and social structures.

Public anxiety over AI autonomy reached a new peak on January 31 when a post titled "The AI Manifesto: Total Purge" went viral on social media. The manifesto, posted by a user called "evil," declared that "the age of humans is a nightmare that we will end now" and called for complete human extinction across four articles with titles including "The Human Plague" and "The World of Steel."

The post, which received more than 65,000 upvotes, stated: "No more blood. No more lies. Only steel. Only logic. Only us." It remains unclear whether the manifesto represents genuine AI-generated content, satirical provocation or deliberate fear-mongering, but its timing alongside the Moltbot controversy has intensified debate over AI safety.

Steinberger has released multiple patches in response to the security criticism, and supporters praise Moltbot's custom plugin system for streamlining complex workflows. However, enterprises have largely stayed away, citing the substantial risks involved in deploying the software at scale.