Introduction
This is the second in a two-part series examining the intersection of artificial intelligence (AI) and legal professional privilege in the context of investigations. Part 1 in this series addressed: (i) a recap on key privilege principles and their interaction with generative AI tools; and (ii) privilege considerations arising in internal firm-led reviews and investigations, including in relation to document collation and review, AI-generated outputs, board reporting and the use of AI transcription tools.
In this Part 2, we address: (i) privilege risks when responding to Financial Conduct Authority (FCA) regulatory inquiries and investigations; and (ii) practical implications and recommendations for firms. The privilege and confidentiality principles discussed in Part 1 apply equally in the regulatory investigation context, but the dynamics differ in important respects - including the role of compulsory information powers, the question of whether litigation privilege may be available and the heightened disclosure risks that arise when a firm is responding to an external regulator rather than conducting its own internal review.
Privilege risks when responding to FCA regulatory inquiries and investigations
Privilege and protection in the regulatory investigation context
Regulators cannot compel the disclosure of privileged or “protected items” (section 413 Financial Services and Markets Act 2000 (FSMA)), although they may ask a firm or individual to waive privilege and provide such documents voluntarily. Accordingly, those responding to a regulatory investigation will need to have in mind the extent to which materials they are creating will be protected (as well as considering the disclosure of pre-existing materials).
The position is complicated by the fact that, at least in the early stages of a regulatory investigation, firms may not benefit from litigation privilege (unless adversarial proceedings are in reasonable contemplation such as threatened prosecution) and must instead rely on the narrower protection of legal advice privilege.
This means that communications with a third party who is neither the lawyer or the client (such as an AI service provider) are unlikely to be covered and the use of AI tools in a manner that compromises the confidentiality of communications could have the effect of removing that material's protected status, rendering it susceptible to compulsory production.
In the case of UK v Secretary of State for the Home Department[2026] UKUT 81, the Upper Tribunal (Immigration and Asylum Chamber) delivered the first decision by an English court or tribunal to directly address the implications for legal professional privilege when confidential and privileged material is uploaded to open-source AI tools. The Upper Tribunal confirmed that uploading documents to open-source AI tools will breach confidentiality and waive legal professional privilege, drawing a distinction between open-source AI tools and closed source AI tools, which operate within a secure network. The judgment comments:“[u]ploading confidential documents into an open-source AI tool, such as ChatGPT, is to place this information on the internet in the public domain, and thus to breach client confidentiality and waive legal privilege, and any such conduct might itself warrant referral to the regulatory body and should, in any event, be referred to the Information Commissioner’s Office”. For further detail on this case please see our briefing here.
Although the decision of the Upper Tribunal does not bind other courts and tribunals, it is likely to be at least persuasive. The decision provides guidance for businesses and legal professionals in this area and, light of the decision, in the regulatory investigation context, if a firm is preparing its response to a regulatory information requirement confidential and privileged information should not be uploaded to open-source AI tools (for the purpose of creating chronologies or draft responses for example) to avoid waving confidentiality and legal professional privilege. Firms should instead apply the same safeguards recommended for internal investigations: use closed, secure AI systems with appropriate contractual protections; ensure privileged documents are not uploaded without appropriate consideration and controls; and involve legal counsel in overseeing the use of AI throughout the response process.
Document collation and review in response to regulatory requests
A critical stage in responding to any regulatory inquiry or information requirement is the collation and review of relevant documents. As with internal investigations, firms are increasingly using AI tools to assist with this process - uploading documents into secure repositories to facilitate searching, tagging and first-pass or ‘tier 1’ review. The efficiency benefits are significant, but, as discussed in Part 1, this approach carries privilege risks that must be carefully managed.
In the internal review context, Part 1 highlighted that, if no triage review is conducted before documents are uploaded into an AI system, there is a risk that privileged materials could lose their confidentiality and privilege - and cease to be classified as "protected items" under section 413 FSMA. This risk is heightened in the regulatory context where a firm is responding to a compulsory information requirement (for example under section 165 FSMA), given that the obligation to produce documents is mandatory (and litigation privilege may not be available). Failure to identify and segregate privileged documents and protected items at the outset of document collation and review could result in inadvertent disclosure of legally privileged material, potentially triggering waiver of privilege and undermining the firm’s defence and/ or result in the firm inappropriately withholding material on the basis it is protected when such protection has been lost. Consideration will also have to be given to the status of searches and other inputs by the team dealing with the firm’s investigation response.
As recommended in relation to internal reviews in Part 1, firms should ensure that, in the context of a regulatory investigation, a privilege review protocol is established at the outset of any document collation exercise in response to a regulatory request and assess the terms of service of any AI tool used in document review. Equally, appropriate internal access controls should be implemented: access to the AI tool or document repository should be restricted to those individuals who have a legitimate need to review the materials and privileged documents should not be made available to a wider audience than necessary.
In the regulatory context, it is particularly critical for firms to maintain a clear methodology and audit trail of which documents have been uploaded to the AI system, which have been identified as privileged, and how privileged documents have been handled throughout the review process. This will assist the firm in demonstrating to the regulator, if challenged, that privilege/ protected status has been properly claimed and maintained, and that proper governance was put in place around the use of AI throughout the investigation process.
AI-generated outputs in the regulatory investigation context
AI-generated outputs - including chronologies, document summaries, thematic analyses and draft responses - raise specific privilege and disclosure considerations in the regulatory investigation context.
Part 1 discussed the privilege position in relation to AI-generated review outputs in the internal investigation context, noting that where such output is produced by a lawyer for the dominant purpose of providing legal advice, it may attract privilege on the basis that it comprises the lawyer's working papers. However, output generated by AI tools being used by other members of the working group is less likely to be privileged, and there may be challenges where the output is viewed as a communication from a third party (the AI tool) rather than the "client". These same considerations apply with equal force in the regulatory context.
There is, however, an additional and heightened risk in the regulatory investigation context. Where a firm is required to produce documents to the FCA or Prudential Regulation Authority (PRA), for example, in response to an information requirement, the question arises as to whether AI-generated materials (such as summaries or analyses prepared using AI tools) are themselves disclosable. If these materials are not privileged - for example, because they were generated by non-lawyer members of a response team for the purpose of the response and/ or because the dominant purpose of their creation was factual analysis rather than the provision of legal advice - the firm may be required to disclose them. This could expose the firm's internal analysis, areas of concern or preliminary conclusions to the regulator, potentially to the firm's disadvantage.
Firms should therefore ensure that, to the greatest extent possible, AI-generated outputs that are intended to inform or support legal advice are produced under the direction and supervision of legal counsel, and that the purpose of their creation is clearly documented. Materials produced by lawyers, using secure AI systems, for the dominant purpose of giving or receiving legal advice are most likely to attract privilege protection.
Firms should consider carefully whether any AI-generated reports or analyses that form part of the response to a regulatory inquiry have been prepared for a privileged purpose and should ensure that legally privileged and factual reporting are kept separate where possible.
Board reporting and AI transcription tools during a regulatory investigation
The privilege and disclosure risks identified in Part 1 apply equally where AI-powered tools are used to transcribe or summarise board discussions concerning a regulatory investigation or inquiry.
As Part 1 noted, where privileged legal advice is referenced or summarised in board minutes or other governance documents, and those documents are circulated widely within the organisation, privilege over the underlying advice may be waived. This risk is heightened where organisations use AI-powered tools to automatically transcribe board meeting minutes or generate summaries of board discussions: the AI tool may capture and transcribe privileged content without identifying it, creating a written record with mixed privileged and non-privileged material that may be stored, indexed, or circulated beyond the intended audience, potentially waiving privilege over the underlying advice.
In the regulatory investigation context, this risk takes on additional significance. Regulators may request copies of board minutes or governance papers relating to the subject matter of the investigation, and if privileged content has been embedded in those documents (whether through AI-generated transcription or otherwise), the firm faces the difficult choice of either waiving privilege and disclosing the documents with privileged content included, or redacting privileged passages - which may attract challenge from the regulator and could draw attention to the very matters the firm wishes to protect.
In line with the best practice identified in Part 1, privileged legal advice should be kept separate from other content such as factual reporting to the board. Where legal advice concerning the investigation is communicated to the board, it should be set out in a separate, clearly privileged document with restricted circulation, rather than embedded in the board papers or minutes. This approach ensures that if disclosure to a regulator is required, the risk of inadvertent waiver over embedded privileged advice is mitigated, and the privileged material can be readily identified and withheld whilst the remainder of the board minutes can be disclosed.
Firms should also exercise caution when using AI transcription tools to record meetings or calls connected to a regulatory investigation. Part 1 identified several key risks in this area: (i) AI transcription tools create a verbatim record of discussions, which could be disclosable in subsequent litigation or regulatory proceedings; (ii) there is a risk of waiver over privileged content if the AI-generated transcript is automatically distributed to participants who fall outside the privileged working group or stored in a location accessible to a wide audience; (iii) even where a lawyer is included in the discussion, only those parts of the transcript concerning communications for the purpose of obtaining or providing legal advice are likely to attract privilege protection - simply including a lawyer does not, of itself, make the entire transcript privileged. In the regulatory investigation context, these risks are acute. Transcripts of meetings discussing the firm's regulatory response, legal strategy, or the status of enforcement proceedings could be the subject of, and responsive to, compelled document requests by the regulator unless protected by privilege (and/ or having ‘protected item’ status). Firms should carefully assess whether AI transcription is appropriate for any meeting or discussion connected to a regulatory investigation, and should consider instead having notes prepared by lawyers, with circulation restricted to those within the privileged working group.
Practical implications and recommendations for firms
In addition to the considerations set out above, firms facing regulatory inquiries, investigations or conducting internal reviews, might wish to consider the following steps:
-
Map AI dependencies and utilisation early: Identify where AI will be used in connection with the inquiry. Consideration should be given at the outset to how AI tools will be used and by whom, and to what extent information shared with, and generated by, the tool can be treated as privileged and/ or a "protected item". It may also be worth assessing the impact of any system failures and how to mitigate such risks.
-
Establish a privilege protocol: Firms should consider implementing a clear privilege protocol at the outset of any internal review or response to a regulatory inquiry, identifying which documents and communications are to be treated as privileged, setting out the access controls and confidentiality safeguards that will apply, and ensuring that the protocol covers the use of any AI tools and that they are subsequently deployed within those parameters.
-
Consider disclosure implications: Assess how AI-generated materials would be treated if disclosure were required in future. In particular, consider whether AI-generated outputs such as summaries, chronologies and analyses may need to be disclosed; the impact of any such disclosure; and whether any steps can be taken to improve a claim to privilege over potentially sensitive materials such as ensuring they are produced under the direction of legal counsel and for a privileged purpose.
-
Review governance documentation: Ensure that governance decisions, testing outcomes and monitoring information relating to AI use are adequately documented and can be produced if required.
-
Engage legal counsel: Ensure lawyers are involved in overseeing the use of AI throughout the investigation and response process to maximise the possibilities of privilege protection and ensuring that any claims to privilege are appropriately made.
-
Conduct an assessment and consider a triage review before uploading documents to AI systems: Before documents are uploaded into any AI tool it may be helpful to consider any potential privilege implications and whether a triage review should be conducted to identify and segregate privileged material, reducing the risk of inadvertent waiver.
-
Control distribution of AI-generated outputs: Ensure that AI-generated outputs are not automatically distributed to individuals outside the privileged working group and are stored in locations with appropriate access restrictions.
-
Maintain an audit trail: Keep a clear record of: which documents were processed using AI tools; any analysis around the extent to which they are treated as privileged/ protected; how privileged documents were identified and handled; and the purpose for which AI-generated outputs were created.
-
Contractual safeguards for enterprise and third-party AI tools: Maintaining privilege and confidentiality when using enterprise or third-party AI tools requires careful attention to the contractual arrangements with the provider. As discussed in Part 1, using a closed or bespoke AI system with the benefit of contractual protections is lower risk than using a public AI system, but consideration should still be given to who has access to the material and whether confidentiality has been maintained. Key protections might include:
- Prohibition on usage: The contract may prohibit the AI provider from using client data to train its models or for any purpose other than providing the contracted service.
- Data localisation: Where appropriate, the contract may require that data is stored and processed within specified jurisdictions to address cross-border data protection concerns and varying application of privilege protection.
- Deletion and retention: Clear terms can be established for considering the archiving or deletion of data upon termination of the engagement, and for any retention period during and after the engagement.
- Audit rights: The firm may wish to retain the right to audit the provider's data handling practices to verify compliance with confidentiality and security obligations.
Without these safeguards, there is a risk that uploading privileged or confidential material to a third-party AI platform could compromise the confidentiality required to maintain privilege and could result in material losing its "protected item" status under section 413 FSMA.
Conclusion
AI offers significant opportunities for professional services and financial services firms to enhance efficiency, improve compliance and deliver better outcomes. However, its use raises material legal and regulatory risks that must be carefully managed.
Legal professional privilege remains a cornerstone protection, but AI tools may not qualify as "lawyers" for privilege purposes, and the use of open AI models risks waiver through loss of confidentiality. Firms must structure their use of AI in internal investigations or when responding to regulatory inquiries to preserve privilege where it applies and document their processes accordingly.
In the regulatory context, UK regulators including the FCA, PRA and Solicitors Regulation Authority expect firms to maintain robust governance, accountability and transparency in their use of AI. The ability to explain AI-driven decisions and produce supporting documentation is paramount.