meta_pixel
Tapesearch Logo
Log in
InsTech - insurance & innovation with Matthew Grant & Robin Merttens

Tobi Schneider, Sector Engagement Lead for Financial Services & FinTech, Edinburgh Futures Institute: Creating a new kind of assurance & insurance framework for AI-related risks (391)

InsTech - insurance & innovation with Matthew Grant & Robin Merttens

InsTech

Entrepreneurship, Business, Investing

4.9 • 51 Ratings

🗓️ 25 January 2026

⏱️ 15 minutes

🧾️ Download transcript

Summary

In this episode, Robin Merttens is joined by Tobi Schneider, Sector Engagement Lead for Financial Services & FinTech at the Edinburgh Futures Institute, to unpack one of the most ambitious research initiatives currently shaping the future of AI risk in insurance. Backed by UKRI and developed in collaboration with AXA Group and three leading universities, the project aims to build a foundational blueprint for how insurers can understand, audit and underwrite emerging AI risks.  Tobi shares why the shift from traditional to generative and agentic AI has outpaced current risk frameworks, leaving insurers exposed to risks that are poorly defined, difficult to monitor and impossible to price using historic loss data. He explains how his team is exploring dynamic underwriting models, parametric solutions and novel assurance techniques like LLM-based judges and automated red teaming, all with the goal of enabling safer, more accountable AI adoption.  Ahead of the Agentic AI Half Day event, hosted in collaboration with AI Risk, Tobi Schneider and Lukasz Szpruch wrote an article The New Frontier: Managing and insuring generative and agentic AI risks, further exploring this topic.   In this conversation, Tobi shares:  Why AI systems that function “correctly” can still produce harmful or costly outcomes  How traditional insurance models fail in the face of opacity, model drift and dynamic learning  What makes AI risk so difficult to price and how parametric triggers can help bridge the gap  Why better assurance leads to better insurance, and how incentives can drive safer AI deployment  How continuous monitoring tools are being developed to audit AI models in real time  What today’s early AI insurance offerings (from the likes of Munich Re and Relm) are actually covering  The role of non-profit research in supporting commercial innovation without commercial bias  What insurers can do now to prepare for an AI-driven future even without historical data  If you like what you’re hearing, please leave us a review on whichever platform you use or contact Robin Merttens on LinkedIn.  Sign up to the InsTech newsletter for a fresh view on the world every Wednesday morning.

Transcript

Click on a timestamp to play from that location

0:00.0

Hello, welcome or welcome back to the Instech podcast, Zoya Here. And this week, Robin is joined by Toby Schneider.

0:16.8

Toby is a leading voice in AI Risk at the Edinburgh Futures Institute, and today's episode

0:22.1

tackles a big question. How do you ensure AI systems that work exactly as intended, but

0:28.4

still produce the wrong outcome? From agentic AI and model drift to assurance and writing,

0:34.2

and why insurers are struggling to price the risk. This is a sharp thought-provoking

0:39.8

conversation about what needs to change if AI adoption is going to scale safely. I hope you enjoy the

0:45.6

conversation. Toby is the sector engagement lead for financial services and FinTech at the Edinburgh Futures Institute.

0:57.9

You were introduced to me by your friends at QBE.

1:00.6

Now, you've been engaged on a piece of research, substantially funded piece of research.

1:06.5

Tell us what problem you're trying to solve with that research.

1:09.5

Yeah, good afternoon, everyone.

1:11.5

So we just started a fairly large resource project between three universities, Edinburgh, Warwick, and Oxford and AXA Group.

1:20.0

I actually seen a picture of our stakeholder within AXA earlier on the slide, where we essentially look at AI insurance and assurance. So we heard a lot about

1:29.2

risks just now. We think that moving from traditional risk to Chen AI, to traditional AI to

1:35.3

a Chen AI, to a Chen AI, the risk profile has changed quite significantly. It's quite material.

1:41.4

And what concerns us is that a big part of this risk is not measured,

1:45.6

not managed and unpriced. So what we're trying to set out in partnership with an insurance

1:50.8

company is to develop a research blueprint, if you want, for how can you actually look at

1:56.9

AI risks from an assurance perspective? So how do you audit AI models?

2:01.0

How do you monitor AI models?

2:02.7

But then how can you develop insurance mechanisms to transfer some of the risk into

2:08.1

insurance markets?

...

Please login to see the full transcript.

Disclaimer: The podcast and artwork embedded on this page are from InsTech, and are the property of its owner and not affiliated with or endorsed by Tapesearch.

Generated transcripts are the property of InsTech and are distributed freely under the Fair Use doctrine. Transcripts generated by Tapesearch are not guaranteed to be accurate.

Copyright Š Tapesearch 2026.