States Look to Rein in AI in Insurance

With Congress punting on artificial intelligence regulation during budget negotiations this year, states remain key watchdogs of the technology as it continues to be embraced throughout American society.

One area of particular concern to the states: insurance, a historically state-regulated industry. 

Colorado became the first state to formally regulate the use of AI in insurance in 2023, although other states before then, including California, Connecticut and New York, instructed carriers to show that their models don’t discriminate.

Lawmakers in New Jersey and Pennsylvania pursued similar legislation in 2022 and 2023, respectively, but neither proposal got off the ground.

So far this year, bills aimed at limiting the use of AI in underwriting and claims have been introduced in at least 17 states. Quite a few of those measures have failed, but a number are still pending and at least two have been enacted. They include:

• Connecticut SB 447 by Sen. Saud Anwar (D), which would have barred insurers from using AI to evaluate and determine patient care. The measure died in committee in March.

• Florida SB 794 by Sen. Jennifer Bradley (R), which would have required all claim denials to be approved by a human, not AI alone. The measure was withdrawn from consideration in May.

• Maine SB 531, a pending bill by Sen. Mike Tipping (D), which would prohibit the use of artificial intelligence in the denial of health insurance claims.

• Minnesota HB 2500, a pending bill by Rep. Alex Falconer (D), which would ban AI’s use in prior authorizations, requiring instead that such a task be handled by a human.

• South Carolina SB 443, a pending measure by Sen. Ronnie Sabb (D), which would require licensed doctors to review healthcare decisions made by AI.

• Maryland HB 820, an enacted measure by Del. Terri Hill (D), which among other things stipulates that AI “does not replace the role of a health care provider” in the utilization review process.

• Nebraska LB 77, an enacted measure by Sen. Eliot Bostar (D), barring the use of AI as the “sole basis of a utilization review agent’s decision to deny, delay or modify health care services.

The legislation highlights the central tension of AI’s use in insurance.

Insurers generally like AI because it can drive down the cost of processing claims. For example, a 2022 McKinsey analysis suggested that AI-enabled prior authorization could automate as much as 75% of the manual tasks associated with pre-approvals.

Critics, meanwhile, are concerned about automated-claims processing’s potential to discriminate or make mistakes. A federal class-action lawsuit against UnitedHealthcare in Minnesota alleges that thousands of Medicare beneficiaries were denied care thanks to an AI tool that has a 90% error rate.

Many States Targeting AI in Insurance

Legislation to restrict the use of artificial intelligence in insurance was introduced in at least 17 states this year, according to the LexisNexis® State Net® legislative tracking system. Two of those states enacted such measures.

Physicians Sound Alarm and Major Insurer Pushes Back

Recently the American Medical Association joined the discussion, issuing a press release in February bluntly headlined “Physicians concerned AI increases prior authorization denials.”

“Using AI-enabled tools to automatically deny more and more needed care is not the reform of prior authorization physicians and patients are calling for,” AMA President Bruce A. Scott, MD said in the press release. “Emerging evidence shows that insurers use automated decision-making systems to create systematic batch denials with little or no human review, placing barriers between patients and necessary medical care. Medical decisions must be made by physicians and their patients without interference from unregulated and unsupervised AI technology.”

The press release specifically called out UnitedHealthcare “as the insurer with the most prior authorization hassle.”

In a January statement UnitedHealthcare said: “We ultimately pay 98% of all claims received that are for eligible members, when submitted in a timely manner with complete, non-duplicate information. For the 2% of claims that are not approved, the majority are instances where the services did not meet the benefit criteria established by the plan sponsor, such as the employer, state or Centers for Medicare & Medicaid Services (CMS). Only 0.5% of claims are not approved based on clinical evidence and patient safety.”

In that same statement, the insurer said AI and machine learning “can help make health care work smarter for our customers, members, providers and consumers.”

It added, “Our teams are focused on developing and using tools that empower people. These tools can give better insights, more timely information and better experiences. At the same time, they keep safety, privacy and transparency a priority.”

UnitedHealthcare says it uses AI and machine-learning models safely and responsibly and that the “tools do not replace critical human decision-making.”

Most of Insurance Industry Now Embraces AI 

Whatever the concerns may be about AI’s use in insurance, the practice is not likely to end soon. The National Association of Insurance Commissioners announced in May that its survey of health insurers found that 84% use AI and machine learning in some capacity.

Indeed, Jack Sloan, vice president of global insurance for the software company Appian, wrote on Insurance Thought Leadership.com this year that “Technology such as artificial intelligence will shape the future of insurance...”

Sloan, who used to hold senior operations positions with Farmers Insurance, added that “2025 will be filled with challenges and opportunities for companies to reap more value and insight from the “insurance value chain.” Insurers that embrace AI and stronger data architectures can improve everything from product development, marketing and underwriting, to policy administration, claims processing and customer service. These gains will help insurers create and sustain more value for their shareholders and customers in the year to come.”

In short, the debate in state capitols about AI’s use in insurance is likely just beginning. Expect to hear more about this for years to come.

—By SNCJ Correspondent BRIAN JOSEPH

Visit our webpage to connect with a LexisNexis® State Net® representative and learn how the State Net legislative and regulatory tracking service can help you identify, track, analyze and report on relevant legislative and regulatory developments.