Bill would require human sign‑off and disclosure when insurers use AI in adverse determinations
Loading...
Summary
Rep. Alicia Gregg's HB 14‑06 would bar insurers from using AI as the sole decision‑maker for claim denials and payment reductions, require disclosure when AI is used, and compel recordkeeping for audits; the Insurance Department supports guardrails but insurers warned the bill could duplicate or expand existing regulatory authority and unintentionally slow administrative functions.
The Commerce Committee examined HB 14‑06, a bill that would add AI‑specific safeguards to New Hampshire law for health insurance claims processing.
Sponsor Representative Alicia Gregg told the committee that AI systems can be trained to produce cost‑driven outcomes and that denials or payment reductions should be made by a qualified health care provider, not solely by an algorithm. "When it comes to denying or reducing health care claims, real doctors should make the final call, not algorithms," Gregg said.
Michelle Heaton of the New Hampshire Insurance Department said regulators have been monitoring AI use, issued a 2024 bulletin and are updating utilization‑review rules (INS 1,000 series) to include AI provisions. She described ongoing rulemaking and said the department already has broad authority to request documentation and audit how insurers use AI.
Insurer and trade representatives pushed back on some draft language. Cam Lapine of Cigna and Paula Rogers of AHIP said carriers do not use AI alone to deny claims, and they argued that lines in the bill that would fold payment reductions or "downcoding" into the statutory definition of "adverse determinations" could expand the law from clinical determinations into administrative coding functions, adding cost and delay. Sabrina Dunlap of Anthem said the amended bill was improved from its original form but remained broad in places and might have unintended consequences.
Supporting witnesses included software engineers and patient advocates who said the bill provides transparency and accountability while preserving efficient uses of AI for fraud detection and administrative processing. The committee did not take a vote; agency rulemaking and statutory language remain under active discussion.
What happens next: The Insurance Department and stakeholders will likely continue to refine statutory language to avoid unintended expansion into routine administrative functions while ensuring explicit human oversight of adverse clinical determinations.
