Europe’s insurers will move to clarify cover for AI risks

Insurance is widely thought to provide ‘silent’ cover for losses related to artificial intelligence at present, but insurers will most likely move to clarify AI-related cover going forward, according to experts in the European insurance market. AI fundamentally has the potential to shift and change insurance risk, according to Scott Toland, chief underwriting officer, EMEA, Zurich Insurance. “We need to make sure that we fully understand how AI could come to bear on a particular industry and policyholder,” Toland said in an interview with Commercial Risk Europe. Risk managers and insurance buyers are beginning to ask questions as to whether current insurance arrangements would cover losses caused by AI, Toland said. “We are starting to see growing dialogue around [AI coverage] and it has come up at some of our customer advisory boards. But it is only the beginning of such conversations, and we do not have structured views yet on where and how AI affects them,” he said. “We are all on an evolutionary journey. Two years ago we were not really talking about AI to the extent we are today, so it is an issue that has accelerated quickly, and we need to think about how we respond and provide that clarity to customers,” said Toland. Generally, the insurance market is not applying restrictions to cover for artificial intelligence, according to Gregory Van den Top, head of cyber risk consulting for Marsh Advisory in Europe. “As far as we know, we do not see exclusions for AI right now, but some risks will be amplified. So we expect policies will pick up on risks generated by artificial intelligence. That said, you always need to look at the risk specifically to understand if current policies cover it. For example, you may have coverage issues if your existing policies did not already cover the risks now amplified by AI,” he said. AXA XL is one of only a handful of insurers to offer dedicated AI cover under its cyber insurance policy. Last month the insurer announced the launch of a Gen AI coverage, available by endorsement to its Cyber Risk Connect policy in North America, UK, Europe and Asia. “It has not been designed for every client. It is really for those organisations that are investing in developing and training their own AI solutions or intentionally training large language models,” said Xavier Veyry, AXA XL’s Europe and APAC CEO. According to Veyry, the AI endorsement expands cover in three specific areas: data poisoning (manipulating or contaminating AI training data); non-intentional violations of the EU’s Artificial Intelligence Act; and cover for usage rights, such as infringement of copyright or intellectual property related to data used in machine learning. “We are still in an environment where as an industry we need to keep on innovating and keep on making sure we bring clarity of cover, so our clients are not surprised at the end of the day by a position taken by an insurer following an incident,” said Veyry. While many insurers are using AI to enhance internal processes and services, the industry also needs to think about how insurance policies respond to AI risks, according to Toland. “Largely, AI is not excluded in current policy language, but we will want to ensure clarity for our customers, and that they understand what the policy covers,” he said. “Endorsements can serve to clarify, or limit, or to assign risk to where we think it should appropriately sit. These are challenges that need to be addressed and looked at,” said Toland. AI could have implications for most lines of insurance, not just cyber, according to Toland. “It could be in liability, it could be in property; there are instances in which AI could be part of any risk exposure,” said Toland. Veyry also expects that further clarification of insurance contract wordings will be needed going forward. “I think we are at a good stage now. Will there be further specific developments related to AI? The answer is probably yes. We will keep on gaining knowledge and understanding of the risk, and adjusting our products. That is for sure. But we are looking to bring clarity of cover to our insureds and making sure we are as transparent as possible in our support for customers, which means we also need to develop the appropriate covers,” said Veyry. “For us, what is important, is clarity. We need to make sure as much as possible that policies are clear, and then we need to design the covers that will allow companies to cover the specific risks that are excluded. This is the direction of travel that we have been in for the last couple of years, making sure that we go from silent to affirmative cyber covers,” he said. Van den Top recommends that insurance buyers review their insurance coverages from an AI perspective, and how insurance would respond to certain loss scenarios, and the effect of cyber exclusions, such as those in property policies. “The cyber insurance policy is an important policy [for AI risks], while liability and technology professional indemnity policies can also provide cover. Depending on what your AI does, you may want to consider some of the physical aspects. If you have automated your production processes, for example, you might want to look into your property policy and cyber policy to see if there is cover for those types of scenarios,” said Van den Top. The development of AI is very rapid, according to Van den Top. “AI has attracted a lot of hype and attention, and we are seeing many organisations starting to experiment with artificial intelligence. But the experimentation stage is just beginning, and a lot of rigour will need to be applied to achieve a return on investment for AI. It is very expensive, and there are real ESG concerns, so we need to look carefully at where we use AI and where it brings the most benefit,” he said. Some 89% of risk managers work on AI-related risk management, yet most monitor regulatory developments (34%) or deploy internal policy (29%) concerning AI usage. Only 23% are working closely with IT on the identification of risks and mitigation actions, while just 19% are mapping current use and performing analysis of risks prior to using AI. “From a risk management perspective, we need to get involved more, and better understand AI and how it impacts our businesses. We can do better assessments, and drill down into what AI is, how it works, and what it does for the organisation, to understand the risk,” said Van den Top. According to Van den Top, AI is a risk amplifier. “So far, AI hasn’t created net new risks, but it rather amplifies existing risks, such as cybersecurity or data privacy. For example, with artificial intelligence and social media, misinformation now has an incredible spread. Where misinformation once influenced us on a one-on-one basis, it now affects the entire world. And that changes the risk profile,” he said.

Europe’s insurers will move to clarify cover for AI risks

While the market is not imposing restrictions, some risks are amplified by AI and insurers want to be clear about what is covered

Want to read this article?

Read Next

US workers comp renewals remain competitive

Aon urges firms to heed UK warnings on AI risk

Consilium brings CyberCube’s analytics to clients worldwide

UK unprepared for scale of wildfire risk

Aon raises data centre capacity to $5bn

US workers comp renewals remain competitive

Aon urges firms to heed UK warnings on AI risk

Consilium brings CyberCube’s analytics to clients worldwide

UK unprepared for scale of wildfire risk

Aon raises data centre capacity to $5bn

Want to read this article?

Read Next

US workers comp renewals remain competitive

Aon urges firms to heed UK warnings on AI risk

Consilium brings CyberCube’s analytics to clients worldwide

UK unprepared for scale of wildfire risk

Aon raises data centre capacity to $5bn

Related Articles