Practical AI

The episode contains genuinely useful structural information - the three-layer standards framework (organizational/infrastructure/agentic), the P0-P4 severity grading for red team findings, and the nondeterminism argument against spotless audit reports - but is diluted by biographical backstory, two product ad reads, and a drawn-out health metaphor tangent where the host seeks validation of his own analogy.

“six of the 40 mandatory requirements in AAC one have to do with red teaming”

“every time you, for example, replace the LLM in an agent, it will behave differently. And if you don't take that into account in your governance, your end users will bear the burden of that”

The Benjamin Franklin flywheel (standards + audits + insurance) is a familiar technology-adoption narrative, but the argument that a spotless audit report is actually less trustworthy than one reflecting reality is a genuinely counterintuitive and underappreciated point; most of the episode is practical rather than contrarian or first-principles.

“a spotless audit report is probably not as valuable as a audit report that reflects reality”

“All agentic systems are nondeterministic in nature. That means that they will always, if you put them under the right amount of pressure, be able to be jailbroken.”

Emil Lassen is a genuine practitioner who built a real standards body with paying enterprise clients including publicly-traded companies, not a career thought-leader; however, his background is governance and entrepreneurship rather than deep technical AI, and the conversation occasionally reveals limits in technical depth.

“we've had a three person Y Combinator startup go through this. We've had UiPath that is publicly traded go through this.”

“companies like Eleven Labs, companies like Fin that just got acquired for 3,600,000,000 by Salesforce”

The episode is notably specific by B2B podcast standards - named clients, named auditors, named frameworks, concrete requirement counts, dollar figures, and a defined severity taxonomy - though a few numbers feel garbled in transcription (e.g. '1,005 different scenarios') and some claims lack supporting data.

“Fin that just got acquired for 3,600,000,000 by Salesforce”

“six of the 40 mandatory requirements in AAC one have to do with red teaming”

Daniel asks two legitimately sharp questions - the forcing-function question and the 'what does passing mean' question - but the interview is fundamentally a friendly PR-adjacent chat; he embeds ads for his own company mid-episode, seeks validation of his own metaphor rather than challenging the guest, and never pushes back on a single claim.

“what is the forcing function that is kind of making making companies consider actual implementation of those of those principles rather than having it be a be an aspirational thing?”

“I I love your answer and it was a little bit I I was trying to validate some of my own thinking through that”