FOLLOW US

Key Takeaways

• Ambient AI can turn conversations into draft records, reducing administrative work.
• Lawyers and insurers face confidentiality, consent and accuracy risks when using these systems.
• Existing professional duties and consumer protections still apply when AI assists with decisions.

CALIFORNIA - Artificial intelligence is moving beyond the chat window and into the conversations that shape medical care, legal advice and insurance claims. Known broadly as ambient AI, the technology uses information from its surroundings to provide quicker, real-time assistance with fewer direct instructions.

Its promise is straightforward; less time documenting a conversation and more time participating in it. For lawyers, insurers and their clients, however, the same capability raises a consequential question—what happens when software misunderstands, records privileged communications without consent, stores or shares something sensitive?

As an example of ambient ai working in real-life, back in March 2025, Microsoft announced Dragon Copilot, with capabilities that include capturing patient-clinician conversations, generating clinical documentation and drafting referral letters and after-visit summaries. Those functions illustrate how spoken information can become working documents. Microsoft’s announcement.

Similar capabilities could theoretically help law firms prepare intake summaries, organize witness interviews and identify follow-up questions. An attorney could spend more of a consultation listening while software prepares a draft account for review.

The risks are equally practical. In a hypothetical interview, a client saying they “might have received” a notice could be summarized as having received it. That change could distort a case timeline or influence legal strategy.

Client confidentiality also makes vendor access, storage and secondary uses of recordings material concerns. Recording consent and permission to disclose client information are separate issues that firms would need to evaluate before implementing an ambient ai tool.

California illustrates the recording problem. Penal Code section 632 generally prohibits intentionally recording or electronically eavesdropping on a confidential communication without all parties’ consent, subject to statutory qualifications and exceptions. Calling software an assistant does not resolve whether its operation complies with that law. California Penal Code section 632.

For insurers, potential uses include summarizing claims calls, organizing loss descriptions and preparing adjuster notes. Accurate summaries could reduce repetitive data entry and help claims professionals identify missing information sooner. It can also become discoverable evidence in ligation, thus escalating potential exposure.

The National Association of Insurance Commissioners’ AI model bulletin states that regulated insurance decisions must comply with applicable laws, including protections against unfair discrimination, regardless of the technology used. The bulletin provides a framework for state regulators; it is not itself a single nationwide statute. Its implications extend beyond ambient tools to broader insurer AI systems. NAIC model bulletin.

Medical information adds another layer. Federal guidance identifies transcription vendors and certain cloud providers handling protected health information for covered entities as business associates. Their arrangements generally require contractual safeguards under HIPAA. Those obligations address data handling and do not substitute for evaluating applicable recording laws. HHS guidance.

The practical test for ambient AI in law and insurance will be whether organizations can demonstrate that time saved comes with dependable records, appropriate consent and accountable human review.

{Also see: Ambient Ai In the Physical Security Space