Summary
1. Client AI chats can form part of the documentary record. If relevant and not privileged, they are likely to be discoverable.
The issue
2. Clients are increasingly turning to AI tools like ChatGPT and Copilot to test ideas including on live legal matters. What many may not realise is that their AI searches and prompts may be discoverable in civil proceedings.
3. In New Zealand, AI interactions are likely to be treated like other documents (e.g. emails). If the AI prompts and results are relevant to the dispute, within your control, and not privileged, then they would be discoverable in civil proceedings.
Why does this matter?
4. AI creates a fast, informal, and largely unprotected record of a client’s thinking. Risks include:
4.1. Unfiltered thinking: AI use can produce candid, or speculative material that a party may not commit to a formal document. If discoverable, it may give the opposing party a window into internal strategy and perceptions. For example, a client asking AI, “What is the likely fall out if we terminate this contract?”, may create a discoverable record of the client’s risk evaluation and the process used.
4.2. Volume risk: AI can produce considerable volume quickly, which increases the risk of unhelpful material being generated or explored.
4.3. Privilege risk that gives rise to real exposure:
4.3.1. Legal advice privilege[1] will not apply to communications between a client and AI tool.
4.3.2. Litigation privilege[2] may not apply, especially at an early stage where litigation is not in contemplation or the dominant purpose test is not met.
4.3.3. Privilege may be at risk if confidential material is shared with third party AI providers, depending on the safeguards in place.
5. Practically most client interactions with AI should be assumed not to be privileged.
Practical suggestions
6. Treat AI interactions as discoverable documents like an email.
7. Don’t do AI searches of things that you would not put in an internal memorandum.
8. For potential disputes, involve lawyers early and have an internal client protocol that deals with both emails and AI use.
9. Don’t put sensitive or significant facts into public AI tools.
10. Consider AI discovery proactively: both as a risk to manage, and a useful weapon against other parties, if your own house is in order.
Conclusion
11. Client AI chats are part of the documentary record, and potentially a revealing part. However, if an issue is serious enough to warrant legal advice, obtain it from a lawyer not an AI tool. That is how privilege is preserved and risk is managed.
This publication is intended to provide commentary and general information. It should not be relied upon as legal advice. Formal legal advice should be sought in particular transactions or on matters of interest arising from this publication.
