Shownotes
A vendor ships an AI product with a safety report written by the same team that built it. Forty pages, clean charts, a passing grade — the system passed its own test, in its own lab, against its own definition of safe. Maya and Theo walk through the shape of it: the vendor picks the questions and grades the answers, the edge cases that break it never make the list, and "safe" means whatever the vendor says it means. The hard rule: a test only means something if the tester can afford to fail it. Blind Spot Audit — Blind guy, sharp eyes for flaws. He audits AI workflows. Company of one.
Website: podcasters.spotify.com

