Deploying AI, insuring climate risk present legal challenges

1 Vanderbilt Avenue, New York
McDermott Will and Emery LLP, headquartered at 1 Vanderbilt Avenue in New York (at left), hosted an Insurtech Summit in June 2025.
SL Green Realty Corp.

In early June, McDermott Will & Emery LLP hosted an Insurtech Summit featuring discussions on investment in startup insurtechs and how the insurance industry is contending with increasing climate-related disasters and risks. Preceding the event, Digital Insurance spoke with H. Michael Byrne, partner in the insurance practice at McDermott, about the increasing use of AI in insurance, and his perspective on what the industry and government should do to address climate risk.

This article is from a longer interview and edited for clarity.

What AI applications are creating legal and regulatory issues?

H. Michael Byrne of McDermott Will & Emery
H. Michael Byrne, partner in the insurance practice at McDermott Will & Emery LLP.
© Gittings Photography

Making sure there's a human being who's accountable. Depending on the complexity of the AI,  how many humans you should have involved and where they are. Whether you delegate functions to humans or to an electronic platform, it's all the same thing. It's how you design it. It's what goes into it.

In underwriting, using AI to get to application submissions faster. The risk could be underwriting something you don't want to underwrite because the platform didn't work. 

With speed, there's risk. You have people using social media or texting something on Facebook too fast, and aren't thinking clearly. They start arguing with someone. They say things they shouldn't say because they're not thinking. The same thing can happen with any use of technology -- in advanced AI, machine learning, insurance, when you're underwriting.

Regulators are concerned about implicit bias. A large language model relies on what's already out there. Discrimination and inequality are already out there. 

How do you make sure with that speed and those large language models, that you're not violating the law? They're trying to build guardrails around that and the uses of it in insurance. Mostly, the world's moving too fast for everyone, including regulators. 

How do you implement human oversight without reducing the benefits of AI?

That's an issue for this industry. On top of that, not only making sure it's just being used to protect the company and for the benefit of the company – and not the detriment of consumers or the public, but making sure that you're operating within the law.

What does resilience mean in insurance?

A combination of socialization of risk, hand in hand with some elements of a free market. When you see one house standing after a major storm on the Florida coast, it was built on cement stanchions dug deeply into the earth. The bottom floor can blow away. The entire neighborhood was clear cut or leveled, but that house is standing there. That's resilience. Someone decided to live that close to the ocean, and mitigated the risk. You would insure that property all day long, if you actually went on site and visited.

Also, it's going away from subsidizing the losses, which is what we've been doing. There's a flood, FEMA comes out. People that didn't buy flood insurance are surprised that they're not covered, because they think the government is going to come and save them, which it doesn't. It doesn't have the money. It's insolvent. Resilience is changing from that approach, because we already know what these problems are. If you're going to spend the money, spend it on resilience in building. 

What is needed to build back after climate disasters?

Are people going to build back these towns and all the businesses that were there? Are people going to build back in North Carolina? Where are people going to live if they don't build back? Eventually it's going to be everywhere, if everybody isn't treated like it's everyone's problem. We have to do something about it -- elect people that are going to make tough decisions, and really start to think about this intellectually, and not kick it down the road, which is what we've been doing. Insurance can't save all that.

For reprint and licensing requests for this article, click here.
Property and casualty insurance Artificial intelligence Climate change
MORE FROM DIGITAL INSURANCE