As artificial intelligence continues to transform the legal landscape in the UK, its impact on litigation, and in particular, expert witness reports, requires much consideration, says Owen Stacey, trainee solicitor at Maulin Law.
The 2025 Bond Solon Expert Witness survey suggests that 20% of expert witnesses use a form of AI in their reports and, as AI continues to develop, this number is expected to rise significantly in the coming years.
However, there are several risks associated with the use of AI that experts must understand and mitigate, in order to avoid criticism by the Courts, and in extreme cases, negligence claims.
How does the use of AI impact an expert witness’s indemnity?
One important aspect that any expert must consider from the outset is their indemnity arrangements. Most, if not all, expert witnesses hold indemnity insurance for their medico-legal reporting.
Because the use of AI is still new, the level of coverage within these policies will vary from insurer to insurer. Therefore, it is vital for experts to understand the scope of their cover if they are considering using AI, and proactively discuss this with their insurer before using these tools.
If a policy excludes claims relating to AI tools, experts must carefully consider the risk this could pose. Should experts still wish to proceed, a transparent discussion should be held with any instructing party prior to accepting instructions, to make sure they are fully aware of any AI use, and ensure they have no issues with its use.
Civil Procedure Rules
At present, the Civil Procedure Rules are quiet on the use of AI in expert reports. However, it is vital that all expert witnesses remember the overarching principle from Practice Direction 35 that expert reports should be the “independent product of the expert”.
While this principle is typically used to stress the independence of an expert from their instructing party, its insistence that the report should be the sole work of the expert would also naturally extend to third party technologies such as generative AI.
It is also important to note that amendments to the CPR to address the use of AI in expert witness reports are currently being considered. In February 2026, the Civil Justice Council released an interim report on the use of AI for preparing court documents, and proposed an amendment to Practice Direction 35 to include a requirement for experts to identify what form of AI has been used in their report and the specific tools used to do so.
Therefore, it is likely that we will shortly see changes to the CPR to ensure that experts are open and transparent about any use of AI in their reports.
The use of AI to review enclosures
As people are living longer and dealing with more co-morbidities, the volume and complexity of medical records bundles in clinical negligence claims are increasing, which will require more time and work from experts to review prior to writing their reports. Therefore, experts are going to be tempted to use AI tools to speed up this process.
However, the biggest issue with using AI tools for review of medical records is the risk of confidentiality breaches. Many AI tools typically use the data provided to them by users to train their models and offer no guarantees of confidentiality. Therefore, by inputting medical records into these AI tools, experts could be breaching the patient’s confidentiality.
Even if experts could be sure that the AI tools being used are not using the records for training purposes, which could avoid the confidentiality issues, experts must be cautious with the use of AI to speed up records reviews. This is to ensure that no relevant records are missed as current AI tools are often poor at understanding context, resulting in relevant data being excluded.
Many clinical negligence claims can succeed or fail based on a single record within a bundle, and so, experts would still be expected and required to review all records to ensure that the AI has not missed any relevant information that could alter the outcome of their opinion. Therefore, while AI may speed up the process, there will inevitably still be a duplication of work by the expert to ensure accuracy.
How AI is often used for research
Once experts have reviewed the medical records for a matter, their next step is likely to be spending time to research the core issues further to find relevant literature that can be discussed in their report.
There are an abundance of stories where this has gone wrong when using AI. A known issue is hallucinations, AI providing references or literature that don’t exist and is something that an expert would need to mitigate.
Experts must ensure that the reference provided exists. This seems like a very basic point, but there are several cases, particularly in the legal profession, where non-existent references were referred to in court documentation.
Once it has been established that the reference exists, the expert should go further and ensure that the literature is relevant to the point that is being portrayed, otherwise it risks weakening the strength of their overall report.
One interesting use for AI when drafting a report would be its use as a sounding board for potential arguments that would be raised. AI could be used in this way to allow the expert to identify weaknesses, gaps, or counter arguments that would allow them to further strengthen their report.
But again, careful steps need to be taken to make sure that patient identifiable information is not included in this sounding board to ensure that there is no risk of inadvertently breaching patient confidentiality.
The use of generative AI in writing expert reports
Finally, when it comes to drafting the report, one obvious use for AI would be transcribing the expert’s views. If AI is to be used this way, it is still important to check any transcription for spelling and grammar, to ensure that what is being transcribed still makes sense and conveys the points that the expert wishes to raise in their report.
Outside of transcription, when using generative AI to draft a report, there is the risk that the AI alters the views that the expert is trying to portray, resulting in inconsistent views throughout a report.
These inconsistent views could potentially create serious consequences if they are not identified early. If the AI generated points are not consistent with the expert’s own views, this could derail cases as a case strategy is likely to rely on the expert’s report. This could open the expert up to negligence claims because of the costs incurred from relying on these inconsistencies.
Ultimately, parties, and the Court, are instructing experts because of the knowledge and experience that they possess from their years of practice. Therefore, experts should think hard about using generative AI to write reports because sometimes their understanding will go above and beyond issues that AI is capable of, and therefore, by relying on AI, the strength of the expert’s understanding is lost.
To conclude, the use of AI in expert witness reports is an issue that is only going to become more prevalent as AI tools develop. Therefore, it is important that experts remain aware of the issues and risks that the use of these tools pose, and should keep abreast any updates to the CPR or expert guidance in this rapidly evolving area.
