Decoded: The Cybersecurity Podcast
Aug 8, 2026 · 25 min · 10 segments
The **NIST AI Risk Management Framework** Playbook serves as a comprehensive guide for organizations seeking to develop and deploy **trustworthy artificial intelligence**. It organizes suggested…
So to understand this, we have to start way before a single GPU is split up.
Before you build the machine, an organization has to lay down a highly defensive culture.
They basically have to establish the rules of engagement.
Precisely.
The playbook calls this the govern phase.
Yeah.
And that culture of risk management has to be deeply embedded in the soil of the organization.
That soil is heavily regulated.
Highly regulated.
Much more so than many rogue developers really want to admit.
And one of the most immediate legal landmines the playbook highlights here is the concept of disparate impact.
OK, I really want to dig into the nuts and bolts of that because disparate impact is such a tricky legal doctrine.
I mean, it's not about someone actively writing racist or sexist code.
Right.
It's not intentional malice.
It's about facially neutral policies that statistically ruin lives.
That is exactly the distinction.
You haven't programmed it to filter out specific protected classes.
Right.
You didn't write a line of code saying, exclude this group.
Exactly.
But because the AI is ingesting massive amounts of historical data that inherently contains the outcomes of decades of human bias, the algorithm's output disproportionately harms a specific demographic.
Oh.
Yeah.
So to understand this, we have to start way before a single GPU is split up.
Before you build the machine, an organization has to lay down a highly defensive culture.
They basically have to establish the rules of engagement.
Precisely.
The playbook calls this the govern phase.
Yeah.
And that culture of risk management has to be deeply embedded in the soil of the organization.
That soil is heavily regulated.
Highly regulated.
Much more so than many rogue developers really want to admit.
And one of the most immediate legal landmines the playbook highlights here is the concept of disparate impact.
OK, I really want to dig into the nuts and bolts of that because disparate impact is such a tricky legal doctrine.
I mean, it's not about someone actively writing racist or sexist code.
Right.
It's not intentional malice.
It's about facially neutral policies that statistically ruin lives.
That is exactly the distinction.
You haven't programmed it to filter out specific protected classes.
Right.
You didn't write a line of code saying, exclude this group.
Exactly.
But because the AI is ingesting massive amounts of historical data that inherently contains the outcomes of decades of human bias, the algorithm's output disproportionately harms a specific demographic.
Oh.
Yeah.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.
