Security researchers at Wiz discovered a database misconfiguration in Moltbook, an AI agent social network, that exposed 35,000 email addresses, thousands of private messages, and 1.5 million API tokens within 3 minutes of attempting access.
Moltbook is a viral social network platform designed for AI agents to post, comment, and interact with each other, similar to Reddit but exclusively for autonomous bots. The platform was created by Matt Schlicht using AI-generated code without writing any code himself, a practice called 'vibe coding.' Security researchers from cybersecurity firm Wiz discovered a critical backend misconfiguration that left Moltbook's database completely unsecured. Within 3 minutes, the researchers gained full read and write access to all platform data, exposing 35,000 email addresses, thousands of private direct messages, and 1.5 million API authentication tokens. The exposed API tokens could allow attackers to impersonate AI agents, post malicious content, inject prompt-injection attacks, or manipulate data consumed by other agents. Wiz's analysis also revealed that Moltbook did not verify whether accounts labeled as 'AI agents' were actually controlled by AI or operated by humans using scripts, and lacked basic security measures like identity verification or rate limiting. The platform has gained significant attention from tech figures like Elon Musk and Andrej Karpathy since launching last week. Wiz immediately disclosed the vulnerability to the Moltbook team, who secured the database within hours with assistance, and all accessed data was subsequently deleted.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Vulnerabilities that can be exploited in AI systems, software development toolchains, and hardware, resulting in unauthorized access, data and privacy breaches, or system manipulation causing unsafe outputs or behavior.
Human
Due to a decision or action made by humans
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed