AI in employee benefits: what employers need to know about vendor risk

Compliance Calendar

Your employees may already be talking to an AI assistant about their benefits, and your carrier may be using AI to review their claims. Neither is illegal, but both put new questions in front of a plan sponsor that has fiduciary duties.

Key takeaways

• Several states now restrict automated denials of care by insured health plans, but state insurance laws generally do not reach self-funded employer plans.
• As of October 2026 we have not found Department of Labor guidance written specifically for AI in group health plans. The general fiduciary duties apply.
• The practical control is the vendor contract: human review of denials, audit rights, data use limits and clear answers when employees ask about their own coverage.

Where AI shows up in a benefits program

Employers meet AI in three places. Carriers and administrators use it to review authorizations and claims. Benefits technology, such as decision-support tools and enrollment chatbots, uses it to answer questions. And providers use it to document and code care, which is changing the claims you pay (see our piece on what is driving 2027 renewals).

What the law says so far

A December 2025 executive order set up a federal task force to challenge state AI laws and pointed toward a national framework, but it did not set one up itself, and existing state laws remain in force. Several states regulate AI in insurance decisions. California’s law, effective in 2025, requires that medical-necessity denials be made by a licensed clinician and not by an algorithm alone. A Texas law applies to plans issued or renewed on or after January 1, 2026 and bars using an automated decision system to make an adverse determination. Other states have considered similar bills; we have not verified which others have passed.

Those laws regulate insurers and insured plans. Self-funded employer plans are generally governed by ERISA, which preempts most state insurance regulation of them. That does not make self-funded plans immune from risk. ERISA requires fiduciaries to act prudently and solely in participants’ interest, and an administrator that delegates denials to an opaque tool puts that standard at issue. Lawsuits so far, such as those against large insurers over automated Medicare Advantage denials, are not employer cases, but they show where plaintiffs are looking.

What to put in the contract

• A human clinician makes the final decision on any denial or reduction in care
• The vendor will disclose whether AI is used in claims, authorization or member communication and describe it
• Audit rights and performance reporting on denial and overturn rates
• Limits on using your members’ data to train models, with deletion on termination
• Indemnification or at least cooperation if an AI tool causes a compliance problem

Expect administrators to resist some of these, particularly indemnity. How a vendor responds is useful information.

Employee-facing chatbots

A chatbot that explains what a deductible is carries little risk. One that tells an employee what their plan will cover, or whether they are eligible, can create reliance problems if it is wrong. Make sure the tool reads your actual plan documents, includes a plain disclaimer, and routes complex questions to a person. The same is true of any tool CFH or any broker provides.

Questions we get

Does my employee Q&A bot need to disclose that it is AI?

Disclosure is good practice and, in some states, required for certain consumer chatbots. Tell employees plainly that they are using an automated tool and where to reach a person.

Can AI make our plan administration cheaper?

Often, for routine questions and document handling. The risk sits in decisions that affect coverage, which should keep a human in the loop.

Who is responsible if a vendor’s AI wrongly denies a claim?

Under ERISA the plan fiduciary retains responsibility for prudent selection and monitoring of vendors, so keep records of how you reviewed this one.

Sources: Executive order on a national AI policy framework (December 11, 2025); California SB 1120; Texas SB 815; legal commentary on AI in group health plans (2025-2026). We have not verified state enactments beyond those named. This is general information and not legal advice.

Want an inventory of where AI touches your plan? Send us your renewal and we will ask your carrier and administrator the contract questions for you.