Artificial Intelligence - Consequences for professional indemnity insurers when AI fails to perform
Use of AI in providing services
A great deal has been written about the potential impact to insurance industry of the use of AI in technologies such as autonomous transport, which we have considered. This use of AI in such circumstances is ‘overt’ as it is clear that the technology requires AI to function. However, the issues are potentially more complex when considering professional services firms, such as solicitors, accountants, insurance professionals or surveyors, using AI to provide services to their clients. Clients are often not aware of how AI is being used in order to provide such circumstances with the result that it is ‘hidden’.
Prior to the widespread use of technology and increasing use of AI, loss scenarios for professionals were relatively uncomplicated; for example, according to a study produced by QBE, the most common causes of claims against solicitors historically included a failure to correctly identify their client resulting in instructions being taken from the ‘wrong’ client, or inadequate supervision of junior lawyers. If an accountant missed a filing deadline or an IT consultant failed to update his client’s security software the issues were normally easily attributable to human errors that would expose the professionals’ PI insurers to a claim.
The potential risk exposure changes if the work is actually done by AI, either where the client contracts with the AI provider directly or, in particular, via a professional services firm so it is that firm that still ostensibly provides the ‘service’ to its client.
Providing legal advice, undertaking electronic filing, or using a program that automatically seeks to identify security weaknesses in a customer’s IT systems are all examples of work that is currently being undertaken by AI technology. In these scenarios, where does the liability rest for a failure to give correct advice, missing a deadline or failing to identify a threat that leads to a cyber-attack or data breach? The client would no doubt look to the professional with whom they have a retainer. The professional will seek to argue it should be the software developer, and their PI insurers, that should meet the claim.
The position becomes more complicated where the professionals have amended the software or if the software requires the interaction of the professional or a third party to produce its final work product. Here any number of potential insurers may be implicated and, at least, may be required to meet defence costs for their insured as claims are pursued.
The fact that a number of factors may lead to a loss is nothing new, the difference now is that the widespread use of technology that utilises AI contributes to additional complexity and uncertainty for insureds and insurers when assessing risk and apportioning liability.
Scope of coverage
Currently, we believe insurers are dealing with claims that involve AI through PI and cyber policies on an ad hoc basis. In due course, we anticipate insurers may expand the scope of coverage available under traditional PI type policies so that ultimately AI failures come to be considered to be a ‘standard’ risk. We can envisage a scenario where firms using AI declare that to their insurer and the risk of the technology failing is priced into their ‘usual’ policies.
Another possibility is that new insurance products are created for the developers of AI technology which provide cover for claims by third parties against the technology’s ultimate users in the event of a failure and these types of risk are carved out of PI policies.
The best way to try and prevent gaps in coverage before claims arise is for there to be an open dialogue between insureds, their brokers and insurers on the types and levels of AI insureds are using to service their clients. In reality, the position may remain unclear for some time. There may need to be a number of claims, and related coverage disputes, before parties begin to appreciate where the potential exposures for
PI insurers truly lie and the market can respond.
In any event, developments in policy coverage will need to try and keep pace with those in technology in order to try and keep ahead of the curve on which type of policy will respond and to try and mitigate against unexpected and uninsured losses.
Consequences for professional indemnity insurers when AI fails to perform