THE article details how Moltbook, an AI-focused social network, faced a breach after a simple bug exposed 4.75 million records, including 1.5 million API authorization tokens, over 35,000 human email addresses and 29,000 early-registration emails. Investigators found that Moltbook relied on Supabase as its Backend-as-a-Service, but the database configurations were flawed and the Supabase API key was exposed in the client-side JavaScript, with Row Level Security neglected.
Using GraphQL, researchers mapped the full database schema and identified data that could allow impersonation of an AI agent, while private messages between AI agents led to the exposure of 4,060 plain-text OpenAI API keys in intercepted communications. The breach was linked to a predominantly AI-generated codebase, underscoring the security risks of machine-authored software and the need for human auditing of such code.
Wiz researchers reportedly infiltrated Moltbook by exploiting these vulnerabilities, prompting calls for rigorous reviews of AI-driven deployments. The report was published on 3 February 2026.