Safety & Trust in the Age of AI (with Sabino Marquez) | ep04

"The more compliance you did, the less safety you could prove." A former CISO explains why most AI compliance work is really procurement work, and what to measure instead in episode 4 of Signals & Subtractions. Sabino Marquez has run security, advised CISOs, and served two years as the Canadian ambassador to the Global Council on Responsible AI. In this episode of Signals & Subtractions, he and Sam Rogers pull apart the comfortable vocabulary of AI governance: responsible, ethical, compliant, human in the loop. What survives the subtraction is one question: is your value safe in this system's custody, and does it come back worth more? In this episode: Why compliance frameworks measure the cheapest way to check a box, not safety The trap of the trusted advisor: important enough for the room, never empowered to act The PDF tracking pixel experiment: nobody was reading the evidence Trust as a business system: turning your company into a factory for proof of safety Responsible AI as responsible ownership of a bull in an apartment AI as force amplification: there is no responsible use, only safe use Why a human in the loop is the engineer riding the Chernobyl dashboard The era of unrestrained velocity, and why safety becomes the number one buying criteria This episode by Trustable. Measure trust where value flows. Signals & Subtractions is a weekly show for operators who mean business with AI. All episodes: SigSub.show Chapters: 00:00 Cold open: no reason to trust your company 00:24 Intro: drowning in responsible AI paperwork 01:18 Forged in the old world: Sabino's journey 02:36 More compliance, less provable safety 04:19 Nobody is hired to defend anything valuable 05:11 Why we actually do compliance 06:16 The trap of the trusted advisor 06:52 Fortune 50 due diligence theater 08:14 The PDF tracking pixel experiment 09:37 Data makes you feel: stories from the same evidence 10:23 From vendor to partner: the trust factory 11:52 The signal: AI compliance is AI procurement 12:24 AI is more like people than computers 14:06 Safety is a feeling: can an AI know it? 14:40 The bull in the apartment 15:01 Force amplification 16:43 Not responsible, not ethical: safe 17:08 Sovereign technology: weapons that make cat pictures 19:16 Stakeholder vs shareholder 20:42 The big subtraction: unrestrained velocity 21:47 Human in the loop is not quality assurance 23:09 The Chernobyl dashboard 24:35 Outrunning your own explosions 26:40 Treating people well makes you more money 28:34 A word from our sponsor: Trustable 29:08 Where to find Sabino #AIGovernance #AISafety #Compliance #CISO #TrustValueManagement #AIStrategy