EWJ August 62 2025 web - Journal - Page 62
and took effect on 12 February 202513. It explicitly
states that “Gen AI must not be used to draft or prepare the content of an expert report (or any part of an
expert report) without prior leave of the Court.” This
strict requirement reinforces the Court’s position
on the importance of expert independence, transparency, and accountability in the preparation of
expert evidence.
cost items if technical terms like ‘PC sum’ or
‘retentions’ are not recognised.
6. Interpretation of Unstructured Data
Quantum experts often deal with a wide range of
unstructured information, such as handwritten site
logs, scanned PDFs of current documents, and annotated architectural drawings. Even as AI-powered
OCR programs – such as Adobe Sensei or Rossum continue to improve, they may still overlook marginal
notes, misread comments, or struggle to decipher
scribbles that context-specific. Without expert review,
for example, a scanned variation claim containing
handwritten comments, could be misconstrued and
lead to data loss or incorrect classification.
3. The Guidance for Judicial Office Holders in the
United Kingdom, issued on 14 April 2025, outlines
the challenges associated with AI-generated outputs14.
The guidance makes it clear that judicial office holders are personally responsible for materials issued in
their name, and that legal representatives are professionally obligated to ensure the accuracy and appropriateness of materials submitted to the court or
tribunal. It implies that AI-assisted expert reports will
be treated like any other: the expert must be able to
justify their methodology, including the use of AI tools,
in accordance with established evidentiary rules.
7. Resistance from Legal Stakeholders
Some arbitrators, judges, or opposing experts may be
wary of AI-driven approaches, particularly if the
methodology appears overly technical or lacks a clear
explanation. For instance, there may be pushbacks or
credibility issues if a quantum expert relies solely on
AI-generated dashboards or automated summaries
without walking the tribunal through the creation and
validation process. Where expert roles are subject to
stringent codes of behaviour, such as Codes of
Conduct for Expert Witnesses, this scepticism may be
extreme.
Outside of the above guidelines and practice notes,
there are currently no reported case law decisions in
New Zealand or the United Kingdom specifically
addressing the use of AI-generated expert reports in
formal proceedings.
Nevertheless, across most common law jurisdictions, it
remains a core requirement that an expert witness
must personally take responsibility for the content of
their report and explicitly disclose all facts, assumptions, and materials relied upon. These foundational
principles indicate that substantive reliance on AI in
drafting an expert report, without proper disclosure
and the expert’s critical oversight, would likely be inconsistent with the expert’s duty to the court. Such inconsistency may, in turn, be a relevant factor in
determining the admissibility or evidentiary weight of
the report in formal proceedings.
8. Ethical and Professional Responsibility
Based on the prevailing legal settings, the opinions
offered are the responsibility of the quantum expert,
not the AI tool. Therefore, using generative AI to draft
reports, summarise claims, or value variations does
not absolve the expert from their duty to independently verify and stand by their conclusions. For example, if an AI analysis concludes that the claimed
prolongation costs were overstated by the instructing
party or the opposing expert, the expert has an ethical duty to ensure verifiable data and proper reasoning support this, not simply because the model said
so.
Conclusion - Could the Role of the Expert Witness
be Fully Automated?
Depending on the scope of the quantum expert’s
instruction, their duty is to typically assesses not just
whether a cost exists but whether the price is causally
linked and reasonable on the assumption that it is contractually recoverable. While AI can assist in improving speed, accuracy, and consistency in quantum
expert tasks, its application is limited by the need for
context, judgment, and legal reasoning, which still requires human expertise. Therefore, in my opinion, it
is not practical for AI to fully automate the role of a
quantum expert; however, it could be used to improve
the efficiency, accuracy, and consistency of quantum
expert reports.
Admissibility of AI-generated Content in
Proceedings
Courts have not yet imposed absolute bans on
AI-assisted evidence, but they scrutinise its use under
existing expert‐evidence rules. Some of the directions
and guidance provided by courts in different jurisdictions are briefly explained below.
1. In the Matter of Weber,12 the Surrogate's Court –
USA (2024) ruled that the admissibility of AI-generated evidence should be evaluated through a Frye
hearing to determine whether the methods employed
are generally accepted within the relevant professional
field. The Surrogate's Court also highlighted the duty
of legal counsel to disclose the use of AI in the preparation of expert evidence, given concerns regarding
reliability and transparency.
The expert remains fully accountable for the accuracy,
completeness, and conclusions of their report, including any sections generated with AI assistance. Accordingly, the quantum expert must exercise the
utmost care when using AI tools to ensure that all outputs are properly explained and validated. Failure to
do so may render part or all of the report inadmissible due to insufficient transparency around the
AI-generated content.
2. In New South Wales (NSW), Australia, the
Honourable Justice Brian J. Preston issued a Practice
Note setting out the procedural rules for using generative AI in expert reports, which applies to all proceedings in the NSW Land and Environment Court
EXPERT WITNESS JOURNAL
60
AUGUST/SEPT 2025
