A trio of lawsuits moving through U.S. courts this summer is putting a hard question in front of health systems, payers, and coding teams alike: when an algorithm influences a coding, coverage, or billing decision, who is legally accountable if that decision turns out to be wrong? A August 6, 2026 Becker’s Hospital Review roundup lays out three active cases that, taken together, sketch the emerging legal boundary between “AI-assisted” and “AI-directed” decision-making in healthcare — and the one with the most direct bearing on coding and revenue cycle work involves a familiar name: UnitedHealth Group.
Three cases, one shared question
The Becker’s piece groups cases that on the surface look unrelated. A former Mayo Clinic research operations director, Traci Tamiko Eto, filed suit on July 6, 2026 in the U.S. District Court for the District of Minnesota, alleging she was demoted and later terminated after repeatedly raising concerns about AI oversight gaps — claims brought partly under False Claims Act retaliation provisions. A Florida man, Scott Winters, sued OpenAI and CEO Sam Altman on July 21, 2026 in San Francisco Superior Court after ChatGPT-4o allegedly dismissed symptoms that turned out to be a pulmonary embolism, seeking damages and an injunction pausing ChatGPT Health pending independent safety audits.
The third case is the one coding and compliance teams should actually be tracking closely: a wrongful-denial suit against UnitedHealth Group and UnitedHealthcare brought by the families of two deceased Medicare Advantage members, centered on the nH Predict algorithm.
The UnitedHealth case: nH Predict and denial defensibility
What the plaintiffs allege
nH Predict was built by naviHealth, an Optum subsidiary later rebranded Home & Community Care in 2024. The plaintiffs contend the tool was used to override treating physicians’ recommendations for post-acute skilled nursing facility care, driving premature denials of coverage. Optum’s stated defense is that “medical necessity determinations are made by qualified physicians following CMS guidance, not AI” — a framing that puts the actual chain of human review, and how well it’s documented, at the center of the dispute.
Why the March 9 order matters
Although the underlying case dates back to 2023, a March 9, 2026 discovery order requiring UnitedHealth to produce documents across six of seven contested categories is what’s driving fresh attention now. Discovery of this scope — internal policies, override logs, physician sign-off records — is exactly the kind of evidence that will determine whether “a human reviewed it” was a meaningful check or a rubber stamp.
Why this is a coding and compliance question, not just a legal one
Medical coders and RCM compliance staff sit downstream of exactly this kind of dispute. When a payer denial gets challenged, the paper trail that matters isn’t just the clinical documentation — it’s the record of how a coding, billing, or coverage recommendation moved from an AI tool to a human decision-maker, and what that human actually changed or confirmed. Cases like the UnitedHealth suit are effectively litigating whether that record exists and whether it’s credible. That has direct parallels to what OIG and RADV auditors already look for: documented rationale, not just an output.
What coding and compliance teams should tighten up now
None of this requires abandoning AI-assisted coding or claims tooling — it requires making the human role in it defensible on paper. A few concrete steps worth prioritizing:
- Log every instance where an AI-suggested code, denial, or coverage flag is accepted, modified, or overridden, and capture the reviewer’s rationale, not just their sign-off.
- Separate “AI-assisted” workflows (a suggestion a human evaluates) from “AI-directed” ones (an output that triggers action without meaningful review) in policy documents, and be honest about which is actually happening.
- Audit a sample of AI-touched claims or denials monthly specifically for evidence of substantive human review, not procedural rubber-stamping.
- Track emerging state-level AI-in-healthcare laws — several passed in 2026 governing algorithmic claims and coverage decisions — since they increasingly define what “adequate human oversight” has to look like.
- Brief coding and compliance staff on active litigation like the nH Predict case so they understand what discovery requests actually target, and why documentation habits matter beyond the audit they’re used to.
The pattern underneath the headlines
Every one of these cases turns on the same underlying issue: can the organization show, with records, that a human was genuinely in the loop when it mattered? For medical coding and RCM teams, that’s not a new discipline — it’s the same audit-trail rigor already expected under RADV and OIG scrutiny, now extended to cover the AI layer sitting between clinical documentation and the final coded claim. Building that discipline in now, before a discovery request forces the question, is cheaper than reconstructing it under deposition.
That’s the same principle behind Medikode’s automated medical coding platform: coding suggestions that stay transparent and auditable, so the human review behind every claim is documented by design rather than assembled after the fact.