NYSBA Warns AI Use Could Waive Attorney-Client Privilege

The New York State Bar Association (NYSBA) has issued a warning that integrating generative AI into legal workflows could inadvertently strip clients of their attorney-client privilege. The alert, released as part of the association's Attorney Professionalism Forum, highlights a critical tension between technological efficiency and the fundamental duty of confidentiality.

In a column published in the NYSBA Journal titled "Talking to AI Could Talk You Out of Attorney-Client Privilege," the forum detailed how the use of AI-powered tools—including generative AI, meeting assistants, and transcription services—creates significant legal risks. The primary concern is that when an attorney inputs confidential client data into a third-party AI model, they disclose that information to an external provider. Under current legal interpretations, such a disclosure may constitute a waiver of privilege, as the data is no longer held exclusively within the protected attorney-client relationship.

The Third-Party Disclosure Rule

This risk is rooted in the "third-party disclosure" rule, a cornerstone of legal confidentiality. Traditionally, attorney-client privilege is maintained only when communications remain private between the lawyer and the client. If privileged information is shared with a third party, the privilege is generally waived unless that third party acts as an agent of the attorney or the disclosure is deemed essential for the legal representation.

As law firms increasingly adopt Large Language Models (LLMs) for research and drafting, the definition of a "third party" has become a point of contention. Because most AI tools are hosted by external corporations, uploading a case summary or a client transcript to a cloud-based model may be viewed by courts as a voluntary disclosure to an outside entity, thereby breaking the chain of confidentiality.

Implications for Legal Liability

The consequences of such a waiver are severe. If a court determines that the use of a public or semi-private AI tool has waived privilege, sensitive client communications that were previously protected could become discoverable during litigation. This would allow opposing counsel to access internal strategies, private admissions, and confidential work product.

For the legal industry, this creates a significant liability gap. Attorneys who prioritize the speed of AI-assisted drafting over the security of the data pipeline may face malpractice claims or disciplinary action for failing to protect client secrets. The potential for discoverable leaks compromises the foundation of the legal relationship: the assurance that a client can speak freely without fear of future exposure.

The Path Forward

Legal professionals must now evaluate the specific terms of service and data-handling policies of their AI providers. The focus is shifting toward "closed" or enterprise-grade AI environments that offer contractual guarantees against data training and third-party access, which may help satisfy the requirements of the agency exception.

What remains to be seen is how courts will rule on these specific scenarios as AI-related discovery disputes reach the bench. Until clear judicial precedents are established, the NYSBA's guidance suggests a cautious approach to any tool that requires the transmission of confidential data to an external server.