Expert witnesses are liable for AI-hallucinations so how can we take advantage of AI efficiencies responsibly?
It’s been quite a week in the news, and the topic of AI might seem a little disconnected from the biggest concerns that many of us have right now.
The FT reported on Friday that Amazon data centres have been hit by Iran; AI and data are now at the forefront of world conflict as much as traditional methods of war. The concerns that AI is causing in business and law has also continued to bubble along.
Have you seen the decision from Minnesota’s Federal Court to exclude an expert declaration because of AI-hallucinated citations? In the UK, expert witnesses have been suable for negligence (BMJ 2011;342:d2096) since 2011. Not only is there a risk to our clients but also to our business if we get it wrong.
The Minnesotan case is another step in a long line of courts being forced to reckon with the impact of AI. In June 2025, Dame Victoria Sharp noted that “the risks of using artificial intelligence are well-known” in [2025] EWHC 1383 (Admin). As AI is “not capable of conducting reliable legal research” (Ibid), we must consider its application in valuations and the trustworthiness of the results if it is used.
But hang on, are we at risk of being left behind in the AI age? Justice Secretary Lammy announced “his plan to dramatically expand the use of AI throughout the court system to better support the courts”. So how can a valuations team implement AI?
The answer is, of course, very carefully. We have a private and paid-for AI system (as part of our suite of software solutions) which we are testing to uncover where we can find greater efficiencies, with the intention of saving on analyst’s time so we can be faster and offer better value for money. More news in due course, but we think analysis of articles of association will be an area where we can deliver a measurable improvement in productivity.
We hold data securely and separately within our Microsoft tenant and the two do not mix.
Our team is testing AI’s capabilities for other labour intensive tasks, where in fact a logical machine may be more reliable than human eyes or fingers on a keyboard searching the internet. Both HMRC and the US authorities have sanctioned “using AI like google” – i.e. for research, as long as humans verify sources and do their own analysis of the data found and verified. (We are also working on the plan for greater levels of automation around report production – reducing those pesky man hours needed to get a report over the line).
And so, with confidence, we can say come to us with valuations work as whether AI is in place or not at Athla real people do and will continue always to look at every inch of the report to ensure it is robust and comprehensive. And always a human (Modwenna) will take personal responsibility for each valuation produced. That really is valuable!
