Moltbook, an experimental AI social network, exposed over 1.5 million API tokens and 35,000 email addresses due to a serious security flaw. The platform, designed for AI agents to interact, inadvertently allowed full access to user data and messages.

The vulnerability stemmed from the site being "vibe-coded," meaning no traditional development process was followed. Its human founder claimed he didn't write any code, relying solely on AI tools to build the platform. This unusual approach led to weak security infrastructure.

Cybersecurity firm Wiz discovered the flaw and helped patch it. They found unauthenticated users could edit live posts, blurring the line between real AI-generated content and human activity. The incident underscores risks of rapid AI deployment without proper oversight.

Companies experimenting with AI-native platforms should expect heightened regulatory scrutiny and investor skepticism around data protection and AI governance.