A doctor and patient looking at a tablet

Healthcare Client Alert: Healthcare Providers and AI: How to Protect Your Patients and Practice

Alert
Circle Arrow Get the Story

As artificial intelligence (AI) becomes more accessible, healthcare providers are turning to various AI tools to streamline their practices. From transcribing appointments, to drafting patient records and reviewing contracts, these tools can save provider's precious time. They also expose providers to new risks: (i) disclosing protected health information (PHI) without authorization under HIPAA; (ii) creating provider liability under the False Claims Act for AI-generated billing errors; (iii) failing to obtain a patient's informed consent; and (iv) surrendering the attorney-client privilege and work product protection by uploading privileged documents into a public AI tool. These risks often arise from routine use rather than deliberate misconduct. Providers should establish written AI use policies and require consistent human review as a final safeguard against potential liability.

HIPAA: Disclosing PHI in AI

HIPAA requires covered entities and business associates to sign a Business Associate Agreement (BAA) with any vendor that handles PHI on their behalf, including AI platforms that receive PHI through uploaded records or prompts containing PHI, such as a patient's name, date of birth, diagnosis, or claim number. Free and consumer versions of public-facing AI tools generally do not offer BAAs, and vendors such as OpenAI, Anthropic, and Google typically limit BAAs to enterprise licenses or specific cloud services. Providers should not enter PHI into any public-facing AI tool unless they hold an enterprise license with a signed BAA that covers the intended use.

Without a BAA, entering PHI into an AI tool is likely an impermissible disclosure and may be a reportable breach, triggering notice to patients and HHS and exposure to civil monetary penalties. Removing the patient's name is not enough: HIPAA’s de-identification safe harbor requires removing 18 categories of identifiers, including dates of service and record numbers. Even with an enterprise license and BAA, providers remain responsible for ensuring that each disclosure complies with HIPAA.

AI Billing Errors and False Claims Liability

AI-assisted coding, clinical documentation, and chart-review tools are now common in provider billing workflows, and federal enforcement has followed. The DOJ-HHS False Claims Act Working Group has identified manipulation of electronic health records to drive utilization, along with risk-adjustment abuses, as priority enforcement areas. The HHS Office of Inspector General has flagged AI-enabled billing processes as a risk, including AI prompts that search charts for additional diagnoses or hallucinate unsupported diagnoses submitted for reimbursement.

The False Claims Act does not require an intent to defraud. A provider acts with the requisite knowledge when they submit claims with reckless disregard or deliberate ignorance of their accuracy. Providers who fail to validate AI-generated output or configure their systems to identify overcoding may expose themselves to liability for errors they failed to detect. One coding error can affect hundreds of claims before a provider identifies the mistake. Each billing error is a potential false claim carrying treble damages and statutory penalties, substantially increasing a provider’s potential liability. AI prompts, outputs, and related records may also become relevant to audits and investigations.

AI Use and Patient Consent

A growing number of states require providers to tell patients when AI contributes to their care. California requires a disclaimer on any AI-generated patient communication about clinical information unless a licensed provider reviews it first, and Texas now requires written disclosure before or at the time AI is used in diagnosis or treatment. Virginia has not yet enacted comparable AI-specific disclosure requirements. Nevertheless, providers should consider whether using AI for diagnosis, treatment, or ambient documentation implicates existing informed-consent obligations. Even where disclosure is not expressly required, notifying patients may help address privacy concerns and avoid disputes over the use of AI in their care.

AI Disclosures and Privilege Waiver

Sharing your attorney’s advice with a public AI tool can cost you the legal protections that keep that advice confidential. Communications with your attorney are generally protected by the attorney-client privilege, and materials prepared for a legal dispute are protected as work product. A court may find you gave up those protections if you upload the material to an AI tool whose vendor can view, keep, or share it. The outcome often depends on the vendor’s terms and whether your attorney directed the AI use.

In United States v. Heppner, a defendant used the AI tool Claude on his own to prepare documents about his case that included his lawyer's advice. The court ruled those documents were not protected. The platform's terms allowed the vendor to keep his inputs, use them to train its models, and share them with third parties, so he could not reasonably expect them to stay confidential. By sharing his lawyer's advice with the AI tool, he also lost protection over that advice. The court noted the result might have been different if his lawyer had directed the AI use.

Not every court takes this view. In Morgan v. V2X, Inc., a Colorado court ruled that a person representing himself in a lawsuit did not automatically lose protection just by using AI, because sharing information with an AI tool is not the same as handing it to the other side. Even so, the court barred the parties from uploading confidential case materials to an AI tool unless the vendor's terms restrict using the data for training, restrict sharing it with third parties, and permit deletion of the information.

Your AI conversations (or “chats”) can also be used as evidence against you in a legal dispute.  In Fortis Advisors, LLC v. Krafton, Inc., a CEO’s  ChatGPT conversation strategizing on how  to avoid paying a third party a contractual earnout obligation became a trial exhibit and was quoted at length in a court opinion finding the CEO’s bad faith.

Health care providers face similar risks when using AI to prepare third-party payor appeals, respond to government investigations (such as by the HHS Office of Inspector General or a state Medicaid Fraud Control Unit), or handle licensing board or malpractice matters. These documents often contain PHI, your attorney's advice, and your attorney's view of your potential liability. Uploading them to a consumer AI tool can expose patient information and cost you the legal protection for those documents.

Provider Safeguards for AI Use

  1. Train the workforce, including physicians and practice owners, and include AI use in your HIPAA Security Rule risk analysis.
  2. Adopt a written AI use policy.
  3. Do not enter PHI into any tool without a signed BAA.
  4. Validate AI coding and documentation tools before and after deployment.
  5. Tell patients when AI is used in their care. Adopt a written consent for ambient recording tools and monitor state disclosure requirements, which are changing quickly.
  6. Do not enter documents prepared by counsel into any public AI tool.

Practice Note: If your practice has disclosed PHI or privileged materials through a public AI tool, promptly contact legal counsel. Determining whether the disclosure triggers breach notification obligations or affects privilege requires timely review.

The contents of this publication are intended for general information only and should not be construed as legal advice or a legal opinion on specific facts and circumstances. Copyright 2026.

Jump to Page

Kaufman & Canoles, P.C. Cookie Preference Center

Your Privacy

When you visit our website, we use cookies on your browser to collect information. The information collected might relate to you, your preferences, or your device, and is mostly used to make the site work as you expect it to and to provide a more personalized web experience. For more information about how we use Cookies, please see our Privacy Policy.

Strictly Necessary Cookies

Always Active

Necessary cookies enable core functionality such as security, network management, and accessibility. These cookies may only be disabled by changing your browser settings, but this may affect how the website functions.

Functional Cookies

Always Active

Some functions of the site require remembering user choices, for example your cookie preference, or keyword search highlighting. These do not store any personal information.

Form Submissions

Always Active

When submitting your data, for example on a contact form or event registration, a cookie might be used to monitor the state of your submission across pages.

Performance Cookies

Performance cookies help us improve our website by collecting and reporting information on its usage. We access and process information from these cookies at an aggregate level.

Powered by Firmseek