AI in the High Court: New Rules for the Use of Generative AI in Court Documents
Generative artificial intelligence is already changing how legal work is carried out.
Tools capable of drafting, summarising, researching and analysing information can create significant efficiencies. But when AI-generated material makes its way into court proceedings, accuracy is not optional.
From 1 September 2026, a new High Court Practice Direction will set clear expectations around the responsible use of generative artificial intelligence (“GenAI”) in Court Documents.
AI can assist with legal work. It cannot replace human responsibility for it.
What is changing?
The Practice Direction applies broadly to documents used in civil proceedings before the High Court, including:
- Pleadings;
- Legal submissions;
- Affidavits;
- Witness statements; and
- Expert reports.
It also applies not just to solicitors and barristers, but to parties, witnesses, experts and other persons involved in proceedings.
Importantly, the High Court has not prohibited the use of generative AI.
In fact, the Court expressly acknowledges that GenAI may help improve the efficient conduct of litigation, reduce costs and enhance access to justice.
However, the use of AI does not reduce any existing professional or legal obligation owed to the Court. Anyone responsible for a Court Document remains responsible for what that document contains.
The problem with relying on AI without verification
One of the most significant concerns surrounding generative AI is that an answer can appear convincing while being completely wrong.
The new guidance specifically identifies risks including:
- Fabricated cases, legislation, academic references or other authorities;
- Incorrect summaries of genuine legal sources;
- Factual errors;
- Biased or incomplete information;
- Confidentiality and privacy risks; and
- Potential loss of legal privilege.
These risks are particularly serious in litigation.
A fictitious case citation in an internal draft is a problem. A fictitious case citation submitted to the High Court can potentially mislead the Court, cause unnecessary costs and delay proceedings.
“The AI said so” is not verification
The Practice Direction places a clear emphasis on independent human verification.
Where GenAI has contributed wholly or partly to material contained in a Court Document, the person responsible must ensure that the material has been appropriately and independently verified.
Using another AI tool — or simply asking the same tool whether its previous answer is correct — is not sufficient verification.
Court users should instead:
- Fact-check and proofread the material;
- Adapt it to the particular circumstances of the case;
- Confirm that cases, legislation, textbooks and articles actually exist;
- Verify that authorities support the proposition for which they are being cited; and
- Check quotations against their original source.
That distinction is important.
GenAI may be useful as an assistant, but the High Court is making clear that it cannot become the final authority on whether information submitted to the Court is correct.
Do lawyers have to disclose that they used AI?
Not automatically in every case.
Subject to the specific rules applying to witness evidence and expert reports, using GenAI to assist with the preparation of a Court Document does not, by itself, require disclosure.
However, the Court may direct a party to disclose whether GenAI was used and, where it was, the nature and extent of that use.
Anyone using GenAI should therefore be in a position to explain how it was used if asked.
There is also an obligation to disclose any limitation affecting the person's ability to independently verify GenAI-generated material contained in a Court Document.
New declarations for affidavits and witness statements
The Practice Direction goes further where AI could interfere directly with a person's evidence.
A witness statement or affidavit intended to stand as evidence must reflect the individual's own knowledge, recollection and evidence.
Unless the Court directs otherwise, affidavits and witness statements will be required to include a declaration confirming that GenAI has not been used to generate or alter the substantive evidence, including by embellishing, strengthening, diluting or rephrasing it.
Administrative uses such as formatting or spell-checking do not prevent the declaration from being made.
Using technology to tidy the presentation of evidence is one thing. Allowing an AI tool to reshape what a witness says happened is another.
Expert reports will also be affected
Expert evidence receives particular attention under the new rules.
An expert report must reflect the expert's own expertise, reasoning and opinion.
Reports must contain a declaration addressing the use of GenAI, and where AI has materially been used in drafting, analysing, interpreting or generating substantive content, the nature and extent of that use must be disclosed sufficiently for the Court and the parties to assess the reliability and independence of the opinion.
The principle is clear: an expert cannot outsource their expert reasoning to a machine.
Confidentiality and legal privilege cannot be overlooked
Accuracy is only part of the risk.
Lawyers and litigants must also consider what information is being entered into an AI system.
Depending on the tool being used, information submitted to an AI platform may be retained, processed or disclosed in ways the user does not immediately appreciate.
The Practice Direction specifically warns that privileged information entered into non-private GenAI systems may risk losing its privileged status.
It also warns against using GenAI with documents or information subject to court orders or undertakings restricting disclosure where doing so could create a risk of unauthorised dissemination.
For law firms and businesses involved in litigation, this makes having an internal AI policy increasingly important.
What happens if the rules are ignored?
The consequences can be significant.
Failure to comply with the Practice Direction may result in measures including:
- Adverse costs orders;
- Court Documents being disregarded or rejected;
- A claim or defence being struck out in appropriate circumstances;
- Personal costs orders against legal practitioners; and
- Referral of a legal practitioner to the Legal Services Regulatory Authority or their professional body.
The Practice Direction also places an obligation on parties to act quickly where inaccuracies are discovered.
If a party discovers an issue with its own Court Documents, the other parties and the Court should be informed without delay and corrected versions supplied.
The wider lesson: AI does not transfer responsibility
The new rules are significant, but the principle behind them is not particularly radical.
Technology changes.
Professional responsibility does not.
A solicitor remains responsible for the legal authorities cited in submissions. A witness remains responsible for their evidence. An expert remains responsible for their opinion. A litigant remains responsible for material placed before the Court.
Generative AI may become an increasingly valuable part of legal practice, but the High Court's new Practice Direction draws an important line.
AI can assist with the work. It cannot assume responsibility for the work.
For legal practitioners, experts, businesses and individuals involved in litigation, the priority should now be ensuring that appropriate AI policies, verification procedures and confidentiality safeguards are in place before the new rules take effect on 1 September 2026.
This article is for general information purposes only and does not constitute legal advice.
