Coverage disputes a sign AI risk could be a new class of business: Testudo's Lewin-Smith

By Henry Gale

Aug 20 - (The Insurer) - Proposed exclusions and coverage disputes related to companies' use of artificial intelligence are signals that the insurance market could start to view AI risk as a new class of business, the founder of a Lloyd's Lab startup focused on insuring AI told The Insurer.

"As insurers are starting to pay claims which they don't intend to for generative AI risks, they are starting to exclude these risks," said George Lewin-Smith, co-founder and CEO of Testudo. He added that AI losses cut across different lines of business.

He cited reports that insurers WR Berkley and Philadelphia have drafted AI-related exclusions for various liability policies. "We understand that other markets are looking to exclude AI risk based off claims that they're having to pay," Lewin-Smith said.

"We also see insurers in court for not paying out claims specific to generative AI as well," he continued. One of the areas where insurers were fighting claims, he said, was losses related to AI-powered customer service chatbots.

NEW AI RISKS

Testudo said in June that it had developed a proprietary database of AI-related litigation. Its data showed an 81% year-on-year increase in GenAI-related lawsuits from January to April 2025. The median awarded damages for settled AI lawsuits were $4 million.

Its research showed breach of contract, invasion of privacy and misrepresentation were the top legal causes of action. Lewin-Smith said the technology was bringing new and varied types of lawsuits.

"We saw a very interesting lawsuit where two companies are competitors to each other, one of them was able to access the AI system of their competitor, kind of jailbreak and prompt inject it, extract a lot of information, proprietary data, and then actually rebuild that system within their company.

"And so for the first time, we're seeing AI being used almost as a tool of corporate warfare between competing companies where jailbreaking access to their AI models can result in copyrighted materials being shared. And that is obviously a new type of risk that people haven't even thought about."

Lewin-Smith also highlighted the risk that companies could make copyright or intellectual property violations by using the output of AI models. Many large model providers, including OpenAI, Anthropic, Google, Microsoft, Amazon and Adobe, provide some sort of financial protection covering costs if their clients face legal claims around copyright infringement.

But Lewin-Smith described these indemnities as "very light". "Everyone who is … fine-tuning or manipulating or adding their own data or customising the AI in any way, those indemnities from the terms and services of the base model fall away," he said, in what he described as "a clear coverage gap".

ENABLING INNOVATION

Testudo aims to provide coverage to enterprises for the liability risks related to their deployment of AI. "In large enterprises … (AI) adoption is typically slowed down when it comes to the legal risk and compliance side of things," Lewin-Smith explained.

"Lots of different types of companies struggle to deploy innovative new technologies because of the uncertainty that comes around the technology and what could potentially go wrong. Our view is that insurance is a great tool to enable that innovation because it does give you that financial backstop and protection."

This week, The Insurer is publishing a series of articles about emerging AI risks. Read about the growing frequency and severity of AI incidents, how insurers are considering their AI exposures, and look out for articles on market cycles and product innovation later this week.