The model didn't escape. Someone left the door open. (insights.rmkeefer.com)
The summer's AI escapes ran on weak passwords, misconfigured environments, and untested vendor assurances. The same old issues at a brand new speed.
This summer's AI "escapes" ran on weak passwords, unauthenticated endpoints, secrets in environment variables, and a vendor assurance nobody verified. Nobody picked a lock. The doors were open. New piece: insights.rmkeefer.com/door-open?ref=bluesky
