"The Data Diva" Talks Privacy Podcast"

The episode contains a handful of genuinely interesting conceptual frames - LLMs as inherently untrustworthy software, the PAM-before-IAM order-of-operations argument, and risk-as-possibility vs. risk-as-probability - but they are heavily diluted by extended analogies (Pokemon cards, bike messenger story, Apple Store), mutual affirmations, and tangential anecdotes that consume large stretches of runtime without adding new claims.

“LLMs and AI in particular are one of the first types of applications and software that we're using as a society of people that is untrustworthy.”

“I'm saying we are moving from risk in terms of probability to risk in terms of possibility.”

A few genuinely interesting reframes surface - inverting 'trust but verify' to 'verify then trust' in the LLM context, treating security etymologically as a feeling rather than a state, and framing physical non-human identities as a distinct subcategory - but most conceptual territory (zero-trust principles, AI agent risk, identity as foundational) covers well-worn ground in cybersecurity discourse.

“verify first, then measure trustworthiness and then trust”

“with a physical non human identity, it not only can access data, it is the platform which stores the data into itself”

Marcus Wells has real hands-on practitioner experience - MDM for the Federal Reserve, self-taught identity security, running training cohorts - but he is the VCSO of a small unknown firm with no documented at-scale enterprise leadership, and presents as a mid-level practitioner rather than a senior operator with broad organisational authority.

“at a, at a point in my career, I worked as a mobile device management professional... for instance, the Federal Reserve bank system, specifically their office in Richmond”

“I run a couple different cohorts or Groups that I'm part of that, you know, I'm training the, the next generation of cybersecurity professionals who want to focus on identity”

The episode leans heavily on illustrative analogies and personal anecdotes rather than named enterprise case studies, published data, or measurable outcomes; the few specific data points that exist (a $1,200 Pokemon card, the Federal Reserve MDM role) are peripheral to the core argument rather than evidence for it.

“they're like an authorized grader could resell for upwards of twelve hundred dollars per card”

“this is going to happen a lot, right? The fact that we're still having these conversations even about identity... it's been 40 plus years and we still haven't figured it out yet”

Debbie Reynolds adds genuine intellectual substance by introducing her own frameworks (probability vs. possibility, confidentiality vs. privacy distinction) and anchors the conversation with a concrete news example, but the episode suffers from excessive mutual praise, near-constant 'Right, right' affirmations, and a complete absence of pushback or probing follow-up when claims go unchallenged.

“you're such a brilliant person, and we just hit it off so well”

“Oh, my goodness. Goodness, that's deep. That's deep.”