Conversations on AI Podcast- Episode 11: Ethical Agentic AI

And I think one of the key realizations of where we were a year ago when we started Pivot Labs, Pivot Labs started on the 12th of September, was formed on the 12th of September, 2025.

One of the things we very much realized, the world of large language models and therefore subsequently agents, autonomous systems, autonomous machines was massively on the rise.

And we had already seen some of the impact of these AI agents or AI systems on the internet as it stands today.

The internet we know and love, the internet that has given us so much opportunity to explore knowledge, et cetera, et cetera.

But it's not an internet built for machines.

And I think we've seen more recently quite a big impact of what happens when we release these large language model driven agents onto the internet, so there's no accusations, obviously, no company's names.

But the point being is the fact that the World Wide Web as we know it has not been built for machines.

And so Pippa Labs set out a year ago to say, OK, so what does this new web look like? Is it a new web? Is it an extension of the existing web? And during that time, we were researching quite heavily around what the proposed architectures for an agentic web, a web for agents and humans, for machines and humans, what that actually meant.

Fundamentally, how it works from an infrastructure perspective, how it works from a moral and ethical perspective, et cetera, et cetera.

And so within a month or maybe two months of starting Pepper Labs, Carl and I reached out to Brian and Alice, having been introduced via the startup hub at Northeastern University in London.

And we began a conversation, very fortunately, Brian and Alice had just published this paper, or recently published this paper, AI 3D Ethical Model.

Brian, you will get the name of the paper spot on.

I'm sorry, I'm just using the context and concepts of it, not necessarily the intricacies of the name of the paper.

But the paper was very much around how to build ethical and responsible AI, a process step, if you like, an iterative process.

And it fascinated Carl and I because it was absolutely what we needed to ground ourselves in the decisions of everything that we were doing in building infrastructure for the agent web or

the agentic web.

And so we began that conversation back in November last year, maybe December.

We tested out the model.

We did pick the model a little bit for several months.

And we've kept very much closely aligned with bioanalysis thinking through to, as you note, the impact challenge for the MSCs and the MAs at at Northeastern University, where a couple of weeks ago we posed a pretty interesting question, which I'll let somebody else talk about, maybe Carl or Barn or Alice, but a question that is absolutely a moral, ethical consideration about how machines should act on our behalf and how machines are fair.

the the the web is right now and how fair uh the agentic web could be and should it be and so and so these conversations continue and i and carl and i will be working with brian and i so i'm sure for years to come so that's a very rounded introduction alex i hope that's a good start

Yeah, that makes some sense.

Well, so it's been, like you said, a couple months since our last conversation.

And Carl, I'll kick this one to you.

Is there anything that's happened over the last couple months that have kind of transformed your guys' process or really felt like step forward, step sideways? Like, how's the last couple months been since we last caught up?