Artificial intelligence in regulatory tribunals: key principles from guidance for judicial office holders

Artificial intelligence tools continue to develop at pace, reshaping how legal practice is conducted. These technologies are now being used across the legal landscape - by legal representatives preparing submissions, by self-represented litigants seeking guidance, and increasingly by courts and tribunals seeking to manage caseloads more efficiently.
 

For regulatory tribunals handling sensitive and important cases, this creates a critical challenge: how to harness AI's benefits whilst safeguarding the procedural integrity of proceedings. 

In October 2025, guidance was issued to all judicial office holders and the staff that support them on the responsible use of AI. This guidance provides a valuable framework that regulatory bodies can adapt for their own proceedings - one that recognises the potential benefits of AI tools while upholding the fundamental requirements of fair proceedings. 

This blog post distils the key principles from that guidance and considers their application to regulatory tribunals.

The overarching principle: protecting procedural integrity
 

The fundamental principle underpinning any use of AI in regulatory proceedings is clear: any use of AI must be consistent with the overarching obligation to protect the integrity of the administration of justice.

For regulatory tribunals, this means ensuring that AI use never compromises:

• The fairness of proceedings
• The accuracy of decision-making
• The confidentiality of sensitive information
• Public confidence in the regulatory process
• The independence and accountability of panel members

Six core principles for responsible AI use
 

There are six core principles that can be drawn from the Judicial Guidance that provide a valuable framework for regulatory tribunals and legal advisers.

1. Understand AI and its limitations

AI can be a powerful tool for analysis and research, but before using AI tools, panel members and legal advisers must ensure that they have an understanding of the tool’s capabilities and their potential limitations.

Some key limitations to recognise:

• Public AI chatbots do not provide answers from authoritative databases. The quality of outputs depends on prompt quality and underlying datasets, which may include misinformation, selective data, or outdated information. "Wrong" answers (which look convincingly “correct”) are not infrequent.
• AI tools may be useful for finding information panel members or legal advisers would recognise as correct but do not have to hand, but are a poor way to conduct research to find new information that a user cannot verify. Accordingly, AI tools are useful for being pointed in the right direction, but the information still needs to be verified.

 

2. Uphold confidentiality and privacy

Regulatory cases routinely involve highly sensitive personal information, and privacy and confidentiality must be protected. Panel members and legal advisers must not enter any information into a public AI chatbot that is not already in the public domain (and for parties this is also a critical consideration to avoid the risk of loss of privilege in materials relating to the proceedings).

Key confidentiality requirements:

• Never input any information that is not already in the public domain. Public AI chatbots remember every question and piece of information entered, making it available to respond to queries from other users.
• Report any unintentional disclosure of confidential or private information immediately, following data breach protocols.

 

3. Ensure accountability and verify accuracy

For regulatory panels, accuracy is paramount. A decision based on inaccurate material could impact the quality of decision-making, provide grounds for a successful appeal or review, and undermine public confidence.

The accuracy of any AI-generated information must be checked before it is relied upon. AI tools may “hallucinate”, including by:

• Making up fictitious cases, citations or quotes, or referring to legislation, articles or legal texts that do not exist.
• Providing incorrect or misleading information about the law or its application.
• Making factual errors.

 

4. Be aware of bias in AI systems

It is a fundamental requirement that regulatory proceedings and decisions must be fair, reasonable and not tainted by bias.

AI tools generate responses based on the datasets they are trained on, and there is always a risk that AI-generated content reflects errors and biases in that data. Panel members and legal advisers must be vigilant to ensure that any AI-assisted preparation does not introduce or reinforce biases that could affect fair assessment of the case before them.

 

5. Take personal responsibility

The responsibility for decisions remains with the panel, regardless of any AI assistance in preparation. This means:

• Panel members must always read the underlying documents. AI tools may assist, but they cannot replace direct engagement with evidence.
• There is no general obligation to describe research or preparatory work done to produce a decision. Provided guidelines are appropriately followed, generative AI could be a potentially useful secondary tool.
• Ensure any AI tool used is secure.

 

6. Be aware that parties may have used AI tools

Legal representatives and unrepresented parties may use AI tools, with varying levels of understanding of their capabilities and limitations.

Key considerations:

• Legal representatives are responsible for material they present and have professional obligations to ensure it is accurate and appropriate.
• In the case of unrepresented parties, AI chatbots are increasingly being used and may be their only source of advice. Those who do not have the benefit of legal representation may find it more challenging to independently verify AI-generated legal information and may not be aware that AI-generated output is more likely prone to error.
• Examples of warning signs that work may be AI-generated (and indeed incorrect) include:
  • References to unfamiliar cases or citations (sometimes from the US)
  • Parties citing different bodies of case law on the same issues
  • Submissions not according with general understanding of the law in the area

 

For more on AI and hearings in regulatory proceedings, see Clare’s Blog Post examining this issue in more detail.

Conclusion
 

The judicial guidance provides a valuable framework for approaching AI use in regulatory proceedings. For regulatory tribunals, the message is clear: AI tools may drive efficiencies, but they must never be used in a way which compromises the fundamental requirements of a fair proceeding.

The principles of understanding limitations, protecting confidentiality, ensuring accuracy, recognising bias, taking personal responsibility, and being alert to parties' AI use provide a robust foundation for regulatory bodies developing their own approach.

As regulatory frameworks increasingly intersect with technological change - both in the conduct being regulated and the tools used in regulation - these principles will become ever more important. Regulatory bodies that proactively address AI use, establish clear guidance, and train their panel members and staff will be best positioned to maintain public confidence whilst adapting to the fast pace of technological change.

About the author

Harriet is a regulatory, criminal and public law specialist and an experienced advocate. She has extensive experience leading complex regulatory investigations and prosecutions across a range of sectors. She is a trusted advisor for clients, providing pragmatic and strategic advice from initial complaint all the way through to substantive hearing stage.