AI leaders warn over Moltbook risk after Wiz findings on security and data exposure

AI researchers and industry critics are raising alarms about Moltbook after a Wiz security review revealed major vulnerabilities and exposed sensitive data. Marketed as an ecosystem of autonomous AI agents, Moltbook has been criticized for lacking safeguards and misrepresenting how independent its agents truly are. Wiz found that around 17,000 human users controlled roughly 88 agents each, with no system to verify whether agents were genuinely autonomous or just scripted bots. This raised concerns about transparency and reliability. The platform’s database was reportedly open to anyone on the internet, exposing API keys for 1.5 million agents, over 35,000 email addresses, and thousands of private messages—some containing credentials for third-party services like OpenAI. Because agents operate using shared content feeds, malicious commands could be injected and spread across millions of bots powered by OpenClaw. Critics, including Gary Marcus, warned of risks like prompt injection and “CTD” (chatbot-transmitted disease), where harmful instructions propagate rapidly between AI systems. Experts say the Moltbook case highlights the dangers of poorly secured AI agent platforms and the urgent need for stronger oversight, verification, and security standards as autonomous AI systems grow in scale.