AI-agent social network Moltbook leaks 1.5M agent keys
Moltbook was a fast-growing, largely AI-coded social network for AI agents. A Supabase row-level-security misconfiguration left its production database readable and writable by anyone. Wiz found that the exposure revealed agent API tokens, owner emails and private agent-to-agent messages, and would have let anyone hijack any agent's account. Wiz reported the issue on 31 Jan and it was fully fixed by 1 Feb.
Damage
~1.5M agent API tokens, ~35,000 email addresses, 4,060 private DM threads exposed; no confirmed malicious exploitation reported
Sources
Wiz, second source. The ~1.5M registered agents belonged to only ~17,000 human owners. 404 Media reported the exposure on 31 Jan (US time) and Wiz published its write-up on 2 Feb. This is a conventional security lapse on an AI-agent platform, not AI behavior.
Factor ratings and the score are this index’s judgments by a published formula; facts, dates and figures come from the linked sources. Where a lawsuit is pending, the summary reports allegations, not findings.
