Tea and Tech Podcast

Today, I want to ask a question that keeps coming back to me whenever I hear someone explain that a new artificial intelligence system has been evaluated.
Who taught the people doing the evaluating?
And how do we know their training prepares them for what they’re looking at?
We hear that a model has been tested. That experts have assessed its capabilities. That researchers found a problem and introduced a safeguard.
And I find myself stopping at the same point.
Based on what?

We welcome a fresh perspective. But we also expect to retain control over what happens next.
So today, let’s explore that moment—when an AI does something we didn’t expect—and how we might respond.

AI has a trust problem—and the guest list matters.
This week, artificial intelligence news moved so quickly that almost anything I wrote in the morning risked becoming historical fiction by dinner.
But beneath the individual headlines, one much larger story is taking shape.  
It is not simply a story about artificial intelligence becoming more capable.
It is a story about responsibility ...

There’s been a lot of conversation in the AI community lately about slowing down.
And whenever I hear that phrase, I have to admit, I struggle with it a little.
Because I’m not convinced that “slow down” is actually the thing we’re trying to solve for. ….

We talk about guardrails all the time in artificial intelligence.
Put guardrails around the model.Put guardrails around the data.Put guardrails around what the system can do.
And generally, I'm in favour of them.
But this week, something happened that made me realize there is a question underneath all of that safety language that we don't ask nearly enough:
Who gave the guardrail authority?

In 2024, the AI company Anthropic began an internal operation called Project Panama. Court records later revealed that the company purchased millions of second-hand books in large batches. The bindings were removed, the pages were scanned at industrial scale and the remains were sent for recycling.
 
(The factual backbone is supported by the Bartz v. Anthropic court order⁠, the U.S. Copyright Office’s fair-use guidance⁠, the Washington Post’s Project Panama investigation⁠, the Guardian’s reporting on rare-bookseller concerns⁠, and Canada’s Copyright Act provisions on fair dealing⁠.)

Last season, we spent time exploring RAI - Relational Abundant Intelligence - and the idea that our future with artificial intelligence cannot be built only around capability….It must also be built around relationship.
Around trust, around responsibility….and around the human beings affected by the systems we create.
This season, I want to move that conversation into governance.
Not governance as a pile of policies sitting unread in a shared drive.
Not governance as a department that is brought in at the end of a project to approve a checklist.
And not governance as an attempt to slow innovation down.
I mean governance as the architecture that makes responsible innovation possible.

Relational Abundant Intelligence....
Because when we say Relational Abundant Intelligence, we are making a statement that relationship—not intelligence itself—should be the organizing principle.
And that is a subtle but very profound shift.
Most of the conversation today is framed as a relationship between two separate categories:
Artificial intelligence and human intelligence.  But what if...

I don't worry that AI will become more human. I worry that it will become an increasingly accurate reflection of us. And if that's true... then perhaps the most important governance decision we'll ever make isn't about the technology. It's about the culture we build around it

So maybe the question for this next era is not only, “Can we build more intelligent machines?”
Maybe it is, “Can we build more humane conditions around intelligent machines?”
Can we design systems that reduce unnecessary cognitive load instead of adding to it?
Can we create information environments that help people orient rather than spiral?
Can leaders communicate with enough clarity that people do not have to decode every decision like it is an ancient riddle with budget implications?