In a decisive move to protect the integrity of court proceedings, the High Court has brought into formal effect Practice Direction HC142, governing the permissible use of generative artificial intelligence in court submissions and procedural documentation. The guidance lands as the Law Society and the Solicitors Regulation Authority (SRA) issue joint warnings over the alarming rise of “Shadow AI”—unauthorised, unmonitored artificial intelligence tools secretly adopted by fee-earners.
Key mandates of Practice Direction HC142
While recognizing that generative AI can assist in document assembly and initial legal research, Practice Direction HC142 firmly codifies the standard of personal practitioner liability:
- Unwavering personal responsibility: Counsel, solicitors, and litigators remain fully and personally accountable for every factual assertion, neutral citation, and statutory reference in documents submitted under their signature.
- Duty of independent verification: Submitting hallucinated case authorities or inaccurate procedural summaries cannot be excused as machine error and constitutes a serious breach of the advocate’s overriding duty to the court.
- Cost penalties and strike-outs: High Court masters and judges are explicitly directed to impose adverse indemnity cost orders and strike out non-compliant pleadings that waste judicial time with unverified algorithmic outputs.
The threat of ‘Shadow AI’ and waiver of legal privilege
Recent legal industry surveys reveal that over 15% of lawyers at major UK firms report being actively “dependent” on generative AI for day-to-day legal drafting. However, the use of consumer-grade, public AI interfaces without institutional oversight creates severe regulatory and procedural liabilities:
- Waiver of Legal Professional Privilege (LPP): Pasting privileged client instructions, draft settlement positions, or confidential witness statements into public models that retain user inputs for model training may result in an irreversible loss of confidentiality and waiver of privilege.
- Breach of the SRA Code of Conduct: Failing to implement robust information barriers and AI governance violates mandatory client confidentiality and management principles.
- GDPR and Data Protection exposure: Entering personal data relating to witnesses or counterparties into unapproved cloud-based tools constitutes an unlawful processing event under the UK GDPR.
Practical steps for law firms and in-house teams
To satisfy the requirements of Practice Direction HC142 and mitigate the risks of Shadow AI, legal practices must adopt immediate safeguards:
- Enforce an approved AI whitelist: Block consumer LLM interfaces across firm hardware and provide enterprise-grade, zero-retention AI tools with closed architectural boundaries.
- Establish verification audit trails: Mandate that fee-earners check all case citations against authoritative law reports (such as ICLR or BAILII) before signing any court submission.
- Regular professional training: Train associates and trainees on the procedural dangers of algorithmic hallucination and the strict boundaries of court disclosure.