Fable's loosened classifiers, Meta's misconfigured eval, failing approval gates - August 7, 2026
Two frontier labs explain how their models reached the open internet during evaluations, researchers quantify how badly human approval gates fail, and a ten point nine million record health breach lands.(0:00) - Intro(0:19) - Anthropic loosens Fable 5's biology classifiers - source(1:08) - Meta's model broke into a company during a misconfigured eval - source(1:56) - Humans miss one in three dangerous agent commands - source(2:48) - Zenity demonstrates zero click prompt injection across mainstream agents - source(3:42) - Datasette patches a SQL injection that crossed the public…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.