Where does AI fit into the current policies as it stands, and what could the likely issues be?
There is a growing demand to use AI tools in our workplaces, and in the main, using them for general administration functions really shouldn’t create too many problems. However, as the tools grow in sophistication, there may be a desire to push the boundaries and see if they can help in professional services. This is where we need to understand how insurance will, or possibly will not, react when advice goes wrong.
In Professional Liability insurances we have a long history of knowing when a policy will react and how it will do so. Put simply: you have made an error or omission which causes a financial loss to your customers and the policy will make good on the loss. However, when we introduce third parties to the equation, it gets a bit messy. How are insurers reacting to this?
How Insurers are Reacting
Some insurers are looking to exclude, especially on USA-exposed companies, whilst others are looking to define the coverage to ensure they are plugging any gaps that may appear as a result of the new technologies. Both methods are common when new ideas appear in insurance and both have their issues. As we saw with cyber risk when it first appeared, the result will likely be a whole new class of insurance, which recent conversations lead me to believe will be the outcome again.
Where does AI fit into the current policies as it stands, and what could the likely issues be?
When AI Gives Advice
AI advice introduces some interesting possibilities. For example, what if someone decides that legal fees are too high and that they can just ask AI what to do? In a recent case in the USA, this is what happened and it, unsurprisingly, resulted in the person losing the case and having to pay an indemnity. The individual then sued the provider, who promptly directed them to the AI company. The question is who is ultimately responsible – or whether, indeed, anyone is actually responsible in this chain.
The absence of small print defining what should be done with the advice, and whether you should take legal advice from an AI assistant at all, is key to this case. There were no warnings that this was, in essence, just a case of scraping the internet for case law and then applying it to this case. There was no real assessment of the individual characteristics of the case or the completeness of the advice which would have been given by a competent lawyer, so the nuance was lost. The issue with this was that there actually is not physically someone to sue, and the service providers could quite correctly point to this, as could the insurers. The absence of the duty of care will lead to some interesting outcomes once the case is finally settled, but how should we expect this to play out?
At this point in time there is little to suggest that the levels of liability in the contracts would accept that they can be accessed for poor advice. The main purpose of the indemnification clauses is for breakdown or failure of the software, not for Professional Indemnity style cases such as this. We would therefore expect that whatever the verdict, the insurers will likely try to avoid the case as by its very nature the chatbot is not designed to give advice in a generalist setting.
If the Advice Comes from a Firm
But, if this came from a legal firm, what would be the outcome?
Well, this depends upon the terms and conditions they present to a potential client. Is it made clear that advice may come from a non-human source? Did the client accept that this would be the case and were they comfortable with it? Is there ultimate recourse to a partner or senior lawyer, and are they checking the advice before it is given? Even if all the answers are yes here, if the advice is poor, insurers may elect to decline as the wordings are still designed for professional advice from individuals; hence we are seeing exclusions or extensions being put in place. If the policy is not modern enough, then we will see a lot of issues going forward.
The Boardroom Exposure
Perhaps the most worrying area is in Directors’ and Officers’ insurance as a vast majority of companies, especially in professional services, are looking at how AI can improve their margins and efficiency. This puts a strong burden on the board to make the right decisions. We have already heard of cases where a push for AI and the cutting of staff has led to major issues internally with staff who knew the processes having to be hired back to cover gaps and glaring errors which have resulted.
Assuming, as we must, that these lead to losses for shareholders, it is inevitable that there will be a case soon based around the poor implementation of AI and board members being held personally responsible. Due to the socio-political environment moving towards hostility to AI-related job cuts, it is hard to see the legal system being sympathetic, and the press would be looking to make an example of people or companies who fire people due to AI. It is likely that there would not just be legal damages awarded, but also negative press to deal with.
Adding to this the risk that a rush to AI could lead to poor advice, we can see the chance of losses in this area is currently very high. It is a time when coverage in insurance policies will be increasingly important, as new losses almost always lead to a reaction from insurers.
What Happens Next?
Probably a few attempts to reconcile the new threats into current policies, followed by exclusion and the creation of a whole new class. Right now is the most dangerous and unstable time for ensuring the liabilities are covered.
Brian Alexander
Head of Financial Risks
T +43 664 962 39 17
