Bipartisan senators demand the top Medicare insurers disclose how their AI denies rehab care

Sens. Richard Blumenthal (D-Conn.) and Josh Hawley (R-Mo.) wrote this week to the three largest private Medicare insurers, UnitedHealth Group, Humana, and CVS Health, demanding internal records on how their artificial-intelligence systems approve and deny rehabilitative care for seniors, the post-acute physical and occupational therapy that follows a hospital stay.

The letters ask for the algorithms, the training data, the human oversight layers, and the appeal records. They extend a Senate Permanent Subcommittee on Investigations inquiry documented in a majority staff report on Medicare Advantage care denials, and they follow earlier congressional questioning of the same insurers on the same subject.

Medicare Advantage is the private-insurance version of Medicare that now covers more than half of all beneficiaries. Insurers are paid a fixed monthly amount per enrollee and are supposed to cover the same benefits as traditional Medicare, including post-hospital rehabilitation. The Senate's interest in this slice is precise: it is the service line where prior-authorization denials directly delay a senior's return to mobility and independence.

The trigger for the new letters is roughly one month old: a Department of Health and Human Services Office of Inspector General report, first reported by STAT+, that found a continuing pattern of denials inside Medicare Advantage, undercutting the insurers' public claims that barriers to care are easing. The OIG is HHS's internal watchdog, and its findings rest on Medicare Advantage encounter and appeal data, not on anecdote.

When a federal watchdog documents a pattern of denials while the same insurers' press releases describe the same period as one of "reducing friction," the gap is what a public records request is built to surface. The American Medical Association, in its July 10 advocacy update, flagged the same service-line concern.

What readers can follow over the next 60 days is the response itself. A bipartisan letter to three named companies, requesting specific documents on a fixed timeline, is a trackable accountability loop. What counts as a substantive answer is also specific: company-by-company disclosures of which AI tools they use in rehabilitative-care decisions, what the tools were trained on, how a human reviewer can override the model, and how many denials were reversed on appeal.

The senators' choice of rehabilitative care is not incidental. Post-acute therapy is the service line where prior authorization has expanded fastest inside Medicare Advantage, and where denials have the most concrete human cost: a senior who cannot get a home health visit stays in a skilled nursing facility, or in a hospital bed, longer than medically necessary.

The three named insurers had publicly committed earlier in 2026 to easing prior-authorization. UnitedHealth Group announced a 30 percent cut in May and said it was leading an industry effort to standardize prior-auth across plans. The Senate letter makes those pledges testable. The companies now have a public, on-the-record obligation to show what the AI actually does, not what the press release says it does.

What readers should watch is narrow. Substantive insurer responses will name tools, datasets, and override rates. Performative responses will point back at the 2026 prior-auth announcements and decline to disclose the AI layer. A third path is a public letter disputing the OIG methodology, which would be its own accountability moment.

The next concrete milestone is the response window. Senators Blumenthal and Hawley have given UnitedHealth, Humana, and CVS a public clock.