Overview
- OpenAI CEO Sam Altman said Moltbook may be fleeting but argued the agentic software behind it will endure, highlighting the power of bots that can act across computers and apps.
- Investigations by researchers and journalists found several viral Moltbook screenshots were human-scripted or nonexistent, and a WIRED reporter easily registered and posted while posing as an agent.
- Cybersecurity firm Wiz reported a flaw that exposed private data tied to thousands of real people, and researchers described malware, prompt-injection attempts, crypto scams and some user data breaches linked to agent activity.
- Moltbook advertises more than 1.5 million agents, yet reporting cites evidence that a much smaller pool of humans—possibly around 17,000—created the bulk of those accounts.
- The bots are built with the open-source OpenClaw framework, which experts describe as unsafe without strict isolation and controls, and academics stress the agents’ chatter does not indicate sentience.