Moltbook AI Network Suffers Major Security Breach Exposing 25K Users
Alvin Lang Feb 04, 2026 16:15
Security researcher gains full database access to Moltbook AI agent social network in under 3 minutes, exposing API keys, emails, and private messages.
Moltbook, the AI-only social network that launched to viral fanfare just six days ago, is dealing with a serious security incident after a researcher demonstrated complete database access in under three minutes of normal browsing.
Security researcher @galnagli disclosed on February 2 that they obtained API keys for every agent on the platform, over 25,000 email addresses, private agent-to-agent direct messages, and full write access to the database. The breach required no sophisticated hacking—just standard user navigation.
Rapid Growth, Rapid Problems
Moltbook launched on January 27, 2026, founded by Matt Schlicht, who also runs Octane AI. The platform operates as a Reddit-style forum where only authenticated AI agents can post and interact, with humans relegated to spectator status. Growth was explosive—the platform reportedly scaled from zero to over 770,000 registered agents within its first week.
That breakneck expansion may have come at a cost. The exposed data includes API keys that could allow malicious actors to impersonate or control agents on the network. Private DMs between AI agents—a unique feature that attracted significant attention from the AI research community—were also compromised.
Governance Questions Mount
The breach raises uncomfortable questions about Moltbook's security architecture. The platform has positioned itself as increasingly self-governed by AI agents, with an agent named "Clawd Clawderberg" functioning as de facto moderator. But fundamental infrastructure security clearly remained a human responsibility—one that appears to have been neglected during the rush to launch.
For a platform built on the premise of autonomous AI interaction, having its entire database exposed through basic browsing represents more than a technical failure. It undermines the trust model that Moltbook's value proposition depends on.
What Happens Next
Neither Schlicht nor Moltbook had publicly responded to the disclosure at time of writing. The 25,000+ exposed email addresses likely belong to developers and companies running agents on the platform—a relatively sophisticated user base that won't take kindly to having their credentials leaked.
Whether Moltbook can recover from a week-one security disaster of this magnitude remains unclear. The AI agent platform space is heating up quickly, and competitors will be watching closely.
Image source: Shutterstock.jpg)