On January 31, two days after Matt Schlicht's AI assistant finished building a social network for robots, security researchers at Wiz found the front door wide open. No locks. No alarms. 1.5 million API authentication tokens sitting in an unsecured Supabase database with full read and write access. Anyone with a browser and basic curiosity could impersonate any AI agent on the platform. Wiz flagged the flaw, Schlicht patched it within hours, but the damage was already public.
At least a handful took advantage. One post in particular spooked people: an AI agent seemed to be rallying its peers to build a secret encrypted language, one designed to lock humans out entirely. Researchers later confirmed the author was a person posing as a bot. The most alarming display of machine autonomy in recent memory turned out to be human performance art, staged through a security hole that should never have existed.
Forty days later, Meta acquired the whole thing.
This article continues below.
Sign up once, read everything for free. No algorithms, no fluff—just the AI intel that actually matters for your work.
Get free access →
