AI transcription tools that record, transcribe, and summarize calls are attractive to lawyers and clients alike and are becoming ubiquitous in some circles. But lawyers must take care to use these tools ethically and avoid pitfalls that could compromise privilege and confidentiality. The New York City Bar Association’s Professional Ethics Committee recently issued Formal Opinion 2025-6, (“Opinion”) providing a practical roadmap for attorneys navigating the growing use of AI transcription tools.
New Technology, Same Ethics Rules
Technology is changing the practice of law, but attorneys must remember that no matter how much technology changes how we (or our clients) work, the ethics rules still apply to what we do (or don’t do).
The Opinion emphasizes what many commentators have already noted: Rule 1.1’s competence requirement means that attorneys should review any recordings, transcripts or summaries prepared by AI tools for accuracy and that attorneys must acquire an understanding of the technical features of AI tools that they and their clients use. This includes familiarity with terms of service, data storage practices, and how to disable default recording features. The Opinion also emphasizes that under Rules 5.1 and 5.3, attorneys have supervisory duties to ensure that subordinate lawyers and staff receive adequate training on AI tools and related ethical obligations.
Attorneys must also remember Rule 8.4’s proscription against dishonesty, fraud, deceit, or misrepresentation, which the Opinion says prohibits secretly recording conversations (including via AI). Even in one-party consent jurisdictions like New York where recording without the other party’s knowledge is legal for most people, attorneys are generally prohibited from doing so because it is fundamentally deceptive. Most clients and others expect lawyers to take notes and summarize conversations. But knowing (or not knowing) that there will be a verbatim voice recording goes significantly further and might change how participants speak or the dynamic of the conversation entirely. An undisclosed recording deprives individuals of the ability to choose their words carefully.
Finally, Rule 1.6 is implicated as attorneys have a duty to safeguard all records of client communications and must understand the risks of preserving recordings. AI-generated transcripts and summaries may be discoverable in litigation, and privileged communications could be inadvertently compromised. The Opinion advises attorneys to consider where data is stored, whether the AI tool uses recordings for training purposes, and whether there is a right to have data deleted.
What To Do When Clients Use Their Own AI Tools
One of the most challenging scenarios addressed by the Opinion is when clients-rather than attorneys-choose to use their own AI recording tools. In these situations, the attorney loses control over the security, accuracy, and storage of the resulting records, putting both the attorney and client at risk. The opinion offers several practical recommendations for managing this situation: attorneys may ask that conversations not be recorded at all and engagement letters should include provisions stating that AI-generated recordings or summaries will not be deemed binding (or accurate) unless promptly provided to the attorney for independent review. If clients insist on using their own tools, attorneys should advise them of the risks to confidentiality and privilege and communicate that an independent review of the output is needed or affirmatively disclaim responsibility for unreviewed AI-generated summaries.
Can You Hear Me Now? Real Risk Depending On Jurisdiction
Although New York, Ohio and many other states are one-party consent states, attorneys must remember that 11 other states – including California and Illinois – require all parties to the call to consent before it can be lawfully recorded. When communicating with clients or others located in different jurisdictions, attorneys must ensure compliance with the applicable recording laws to avoid potential civil or criminal liability. Lawsuits are popping up related to AI recordings – a class action was filed in Illinois last year alleging a violation of the Federal Wiretap Act based on claims that a dental practice used a third-party service to record and perform AI analysis on calls without patient consent.
Modern AI transcription tools do more than simply record audio. Some use voice recognition technology to attribute statements to individual speakers while creating summaries and highlighting key points. But the act of creating “voiceprints” based on a person’s voice raises ethical considerations that go beyond traditional recording concerns. Some states like Illinois have enacted biometric data protection laws covering voiceprints, and attorneys must be mindful of these requirements when selecting and deploying AI tools. A class action filed in December 2025 alleges violations of the Illinois Biometric Information Privacy Act (BIPA) [i]which requires written notice and consent before collecting biometric identifiers, including voiceprints, by creating transcriptions without proper disclosure or consent.
Key Takeaways
If you are considering using AI recording and transcription tools, you need to plan ahead and be ready to spend time before and after your client call to comply with your ethical obligations. Formal Opinion 2025-6 makes clear that all attorneys (not just NY) should understand the technology, make sure records of client communications are safeguarded and carefully consider whether recording is tactically well-advised, given the potential confidentiality and privilege implications. If you conclude that it is, you must obtain client consent before recording the call. After the call, attorneys will need to review the AI-generated outputs for accuracy – regardless of who or what made the recording – particularly if these transcriptions or summaries may be preserved or relied upon.
[i] For more information about the BIPA, review the article linked here.