Issued by Mr Justice David Barniville, President of the High Court · Signed 29 July 2026 · In operation from 1 September 2026
On 29 July 2026, Mr Justice David Barniville, President of the High Court, signed Practice Direction HC 142 — the Practice Direction on the Responsible Use of Generative Artificial Intelligence in Court Documents. It comes into operation on 1 September 2026 and applies to all civil proceedings before the High Court — including cases already underway. Any court document produced from that date onward falls within scope, whether the case itself started before or after 1 September.
Why now
The judiciary has been watching AI "hallucinations" — fabricated cases, invented citations, misquoted authorities — surface in courts worldwide, and Ireland has had its own experience of this in Guerin v O'Doherty. HC 142 is the High Court's formal response.
Who it covers
Everyone involved in proceedings: parties, solicitors, barristers, witnesses, experts, litigants in person, and any third party required to produce documents. It applies to pleadings (the formal documents setting out each side's case), submissions, affidavits (sworn written statements of evidence) and expert reports.
The core rule
Every duty already owed to the court stays fully in place when GenAI is used. Every person responsible for a court document must independently verify any AI-assisted content before it's used — and that verification has to come from checking against the original source, not from asking a second AI tool to confirm the first (para 13).
What verification means in practice (para 18)
- Fact-check and proofread AI output.
- Edit and adapt it to the actual facts of the case.
- Confirm that every case, statute, textbook or article cited genuinely exists and says what it's claimed to say — extracts and quotes checked against the original source.
Disclosure
Routine use of GenAI stays a private workflow choice — disclosure only becomes required when the court specifically directs it, and everyone needs to be ready to answer that question if asked (para 21).
Witnesses and affidavits — the new declaration
Anyone swearing an affidavit or making a witness statement must open the document with this declaration, unless the court directs otherwise (para 27):
"I declare that Generative Artificial Intelligence has not been used for the purposes of generating the substantive content of this [witness statement/affidavit] (including by way or altering, embellishing, strengthening, diluting or rephrasing the evidence set out) and that this [witness statement/affidavit] reflects my personal knowledge, recollection and evidence."
AI used purely for formatting or spell-checking keeps this declaration available as written (para 28). The deponent — the person swearing the affidavit — carries personal responsibility for its truth.
Expert reports
Experts open their report with a parallel declaration that it reflects their own opinion and reasoning. Where GenAI was used for substantive content, that use must be disclosed in enough detail for the court and parties to assess the report's reliability and independence (paras 29–31).
Litigants in person
The same accuracy standard applies equally here. They're pointed to the Courts Service's plain-language guide, Using AI to prepare documents for court (paras 38–41).
Consequences of non-compliance (paras 35–37)
- Adverse cost orders.
- Court documents disregarded or rejected, in whole or part.
- Claims or defences struck out.
- For legal practitioners specifically: referral to the Legal Services Regulatory Authority or their professional body, and personal costs orders.
The bottom line
HC 142 keeps AI available as a litigation tool — the Court explicitly recognises its potential to cut costs and improve access to justice (para 6). What it requires is human verification of everything that reaches the court, clear declarations from witnesses and experts, and accountability when that verification is skipped. Practitioners have until 1 September 2026 to build this into their workflow.
The pattern here — verification, human oversight, disclosure — is the same one now showing up in Article 4 AI-literacy obligations under the EU AI Act: organisations are expected to build these controls into how staff use generative AI, not just react when something goes wrong in court. If your organisation doesn't yet have that internal AI usage policy, that's the gap worth closing now.
Building an internal generative AI usage policy?
This article is general analysis — it is not legal advice and does not account for your organisation's specific circumstances. If you want a view on your AI governance position or how court-facing verification duties map onto your internal policy, that is exactly the kind of work we take on.
Begin in writing →Sources: Practice Direction HC 142 — full text (courts.ie) · Using AI to prepare documents for court — Courts Service guide · Law Society Gazette: High Court expects 'human supervision' of GenAI