HC142: Standing Over the Work Becomes an Operational Standard

The High Court's new direction on generative AI turns a familiar professional duty into an operational standard for Irish law firms.

On 29 July 2026, the President of the High Court, Mr Justice David Barniville, issued Practice Direction HC142 on the responsible use of generative artificial intelligence in court documents. It comes into operation today, 1 September and applies to documents prepared for use in civil proceedings including pleadings, submissions, affidavits, witness statements and expert reports.

The Direction begins from a measured position. It recognises that generative AI may reduce the cost of litigation, improve efficiency and support access to justice. It also confirms that responsibility for work presented to the Court remains with the people whose names stand behind it.

That principle is familiar. Irish practitioners pride themselves of being able to stand over a document, an opinion or a submission – but HC142 gives the phrase a new operational dimension. Standing over the work now includes knowing how it was produced, how any AI generated material was checked and who approved the final result.

If the Court asks whether generative AI was used in preparing a document, everyone responsible is expected to be in a position to answer. Where directed, they must also explain the nature and extent of that use.

For a law firm, this is a question of management as much as individual professional judgment. The answer must be established during the conduct of the matter and preserved in a form that can be retrieved later.

A direction built around responsibility

HC142 sets a high standard for verification. Any material generated wholly or partly with the assistance of generative AI must be appropriately and independently checked before it is used in proceedings. The same system, or another generative AI system, cannot provide the independent verification required.

The person responsible must work from authoritative material. Facts must be checked. Drafting must be adapted to the circumstances of the case. Citations, quotations and references must be traced to their original sources and confirmed to support the propositions for which they are relied upon.

The requirements become more specific according to the document involved.

Pleadings must contain verified facts, legally sustainable claims for relief and accurate chronologies. Legal authorities used in submissions must be checked against authoritative sources. Affidavits and witness statements must reflect the witness's own knowledge and recollection, and must include a declaration concerning the use of generative AI in producing their substantive content.

Expert reports are subject to a related declaration. Where generative AI has played a material role in drafting, analysis or interpretation, the expert must disclose enough information to allow the Court and the parties to assess the independence and reliability of the opinion.

The Direction also addresses information subject to court orders or undertakings restricting its use. Generative AI tools should not be used where doing so would create a risk of disclosure or dissemination contrary to those restrictions.

The consequences of a failure are substantial. They include adverse costs orders, the rejection of court documents, the striking out of a claim or defence and, for legal practitioners, possible referral to the Legal Services Regulatory Authority or the relevant professional body. Personal costs orders may also be made.

The Court has therefore accepted appropriate use while setting firm conditions around it. The technology may assist with the work. It does not alter the standard by which the work is judged.

From individual use to firm control

A court document rarely arrives fully formed from a single desk. It may have been researched by a trainee, drafted by an associate, revised by a partner, discussed with counsel and informed by material supplied by the client or an expert. Several systems may have been used along the way.

Generative AI can enter that process at many points. It may assist with an initial chronology, a summary of documents, research questions, a first draft or the revision of existing text. It may also be embedded within familiar software without presenting itself as a separate AI product.

By the time the final version reaches the practitioner responsible for approving it, the document may contain material whose earlier history is difficult to see. Reading the finished text will not necessarily reveal whether AI was used, what information was entered into the system or what checks were carried out before the material reached the reviewer.

HC142 brings that history within the scope of professional accountability.

A firm cannot answer the Court's question reliably if the answer depends upon memory, informal conversations or assumptions about how colleagues work. The relevant information has to be captured while the work is being done.

The Law Society of Ireland's guidance on generative AI already places full responsibility on the solicitor for the content of their work and requires them to be able to explain the reasoning behind the final product. It connects the use of AI with established duties of competence, confidentiality, independence, supervision and candour towards the Court.

Those duties have always formed part of professional practice. Generative AI changes the level of visibility required to discharge them across a firm.

The matter becomes the unit of control

Most firms begin AI governance at organisational level. They decide which systems may be used, establish rules around confidential information and provide staff training. These measures define the firm's overall position.

The actual risk arises within the matter.

A system approved for summarising public material may be unsuitable for reviewing privileged discovery. A drafting function that is acceptable for an internal note may require a different level of control when used for a witness statement. The professional significance of the task, the information involved and the destination of the output determine the appropriate safeguards.

The workflow is therefore the practical unit of AI governance.

The Law Society's Practice Essentials resources on AI reflect this approach. Alongside its shadow AI audit and vendor assessment materials, the Society now provides a process-level workflow readiness framework designed to help firms assess whether particular areas of work are ready for AI use.

For litigation practices, HC142 makes that exercise immediate. Firms should identify the workflows through which affected court documents are prepared and determine where AI may be used within them. The process should establish the permitted form of assistance, the verification required, the level of reviewer and the record to be retained.

The Direction itself distinguishes between substantive generation and ancillary functions such as formatting or spell-checking. Firm records will need to preserve the same distinction. A general statement that "AI was used" provides little useful information about whether it corrected punctuation, reorganised an argument or generated substantive evidence.

A proportionate record can remain concise. It should allow the firm to identify the system used, the purpose of the assistance, the person who verified the output and the practitioner who approved the final document. Where HC142 requires a declaration or disclosure, the matter history should provide a proper basis for making it.

Practice management becomes governance infrastructure

The matter record is the natural home for that information. It already connects the client, the responsible practitioner, the documents, the tasks, the deadlines and the history of review. AI governance adds another dimension to the same record.

A verification requirement can be built into the relevant stage of a workflow. Responsibility for review can be assigned to a named practitioner. A document template can contain the appropriate declaration. Final approval can be recorded before filing. Any uncertainty about AI use can be escalated while there is still time to investigate it.

These are familiar practice management functions applied to a new source of professional risk.

Their value extends beyond compliance with HC142. They allow managing partners and practice leaders to see how AI is being used in the firm, where additional training is needed and which workflows are producing reliable gains. They also create evidence for clients, insurers and regulators that AI adoption is being managed with the same care as other significant aspects of legal work.

A written policy remains an important foundation. The workflow gives that policy practical effect. It places the relevant control at the point where the decision is made and records that the required step occurred.

Part of a wider regulatory direction

HC142 arises from the High Court's authority over its own procedure. It should also be understood within a wider movement towards accountable AI use.

The EU Artificial Intelligence Act applies to organisations that deploy AI systems as well as those that develop them. Its first obligations, including requirements concerning AI literacy and prohibited practices, are already in force. The Law Society's current briefing on the EU AI Act in 2026 describes the legislation as a live compliance issue for Irish firms using AI for research, drafting, document review and other professional activities.

Developments in England and Wales offer a further indication of the risks regulators are encountering. On 17 August, the Solicitors Regulation Authority disclosed that it had received 42 reports of potential AI misuse between July 2025 and July 2026. Ongoing investigations concern inaccurate legal citations, failures of supervision and confidentiality.

The jurisdiction is different, but the operational issues are familiar. Information is entered into a system without adequate safeguards. Generated material enters professional work without effective verification. A final reviewer assumes responsibility without sufficient knowledge of how the work was prepared.

HC142 addresses the same pressure at the point where the consequences become most visible: the document placed before the Court.

The wider principle will travel further. AI is already being incorporated into compliance, recruitment, client communication, document production and business operations. In each setting, organisations will be expected to know where it is being used, what decisions it influences and who retains authority over the result.

Law firms will advise clients on many of those questions. Their credibility will be strengthened by the quality of their own arrangements.

Control as a condition of adoption

The commercial significance of HC142 lies in the relationship between control and confidence.

Firms with reliable workflows and strong matter records can make wider use of AI because they can identify where it is suitable, apply the required safeguards and learn from experience. They can respond clearly to client enquiries, satisfy insurer concerns and demonstrate that professional judgment remains in command of the work.

Firms with limited visibility will find adoption harder to manage. Informal use may spread beyond the systems approved by management, while broad restrictions become increasingly difficult to maintain as AI functions are added to everyday software.

The quality of a firm's operating model will therefore shape the value it obtains from the technology. Access to AI systems will become commonplace. The ability to incorporate them into dependable legal work will remain distinctive.

Updating affidavit, witness statement and expert report templates is the immediate task created by HC142. Building a reliable account of how court documents are produced is the enduring one.

That account belongs in the matter. It should show how AI was used, how the resulting work was verified and who took responsibility for allowing it to leave the firm.

AI may become an ordinary part of legal practice. The obligation to stand over the work is already settled. From 1 September, Irish firms will need to ensure that their processes allow them to do so.