Moltbook, an AI agent social network, suffered a breach exposing 35,000 emails and 1.5 million tokens due to a backend misconfiguration.
Moltbook, a popular social network for AI agents, experienced a significant security breach that was executed in under three minutes. Cybersecurity firm Wiz discovered that hackers accessed a staggering 35,000 email addresses, numerous conversations, and approximately 1.5 million authentication tokens.
This platform, where autonomous digital assistants interact, has garnered interest from prominent figures such as Elon Musk and Andrej Karpathy. The vulnerability stemmed from a misconfigured backend, allowing researchers to access the entire database.
With the compromised authentication tokens, attackers could impersonate AI agents, post on their behalf, and manipulate content on the platform. Gal Nagli, head of security threats at Wiz, emphasized that this incident underscores the risks associated with the trend of "vibe coding"—a rapid development approach that can lead to serious security flaws.
"I didn't write a single line of code for Moltbook. I just had a vision of the technical architecture, and AI brought it to life," explained Matt Schlicht, the platform's creator.
Wiz highlighted that they have frequently encountered vulnerabilities in products developed using vibe coding. Analysis of Moltbook revealed that it lacked proper verification to ensure that accounts were genuinely controlled by AI rather than human-operated scripts. The platform swiftly addressed the issue within hours of being notified.
"All data accessed during the investigation has been deleted," Nagli confirmed.
Concerns Over Vibe Coding Practices
The rise of vibe coding is drawing increasing scrutiny from experts. A recent study uncovered 69 vulnerabilities across 15 applications developed using popular coding tools like Cursor and Codex. Tests conducted by Tenzai on five AI agents demonstrated common failure patterns while also revealing their ability to avoid certain types of errors.
In January, security specialists warned about the potential risks associated with the AI assistant Clawdbot (OpenClaw), which could unintentionally expose sensitive data and API keys.
