Generating insurance appeal letters for patients is not clearly UPL — it sits in a gray zone that Claimable, Counterforce Health, and Fight Health Insurance are already operating in without legal challenge. The biggest risk factors for UPHELD specifically are the contingency fee model (which elevates UPL exposure significantly vs. flat fees) and auto-submission without patient review (which removes the "document preparation tool" defense). This is a manageable risk, not a blocker, but deliberate structural choices and a legal opinion from a healthcare/UPL attorney are required before launch.
Insurance claim denials and appeals happen within the insurer's internal process and state insurance department review — they are not lawsuits. Courts consistently distinguish administrative advocacy (helping navigate a bureaucratic process) from practicing law (representing someone in court or giving specific legal advice). This distinction is UPHELD's primary legal shelter.
UPL definitions vary by state but share a core: practicing law means giving legal advice within an attorney-client-like relationship, or representing someone in court/quasi-judicial proceedings. Non-attorneys can help prepare documents and navigate administrative processes as long as they don't give specific legal advice ("you should sue," "your rights under this statute are X"). Patient billing advocates do this every day across all 50 states.
Claimable (covered by Forbes, Bloomberg, NYT, NBC News), Counterforce Health (NIH-funded), and Fight Health Insurance all generate AI insurance appeal letters for patients. None has been prosecuted for UPL. Claimable charges $39.95/appeal and includes clinical evidence and "relevant state and federal laws" in letters. Counterforce calls output "legal-grade appeal letters." Both operate openly with mainstream press coverage.
DoNotPay was fined $193,000 by the FTC (finalized Feb 2025) and faced class action lawsuits — but primarily for false advertising, not UPL. Calling itself "The World's First Robot Lawyer" while delivering sub-lawyer quality output triggered state bar investigations AND FTC consumer protection action. The lesson: never claim to be a lawyer or provide "legal services." Frame as document preparation, administrative assistance, or patient advocacy — never legal representation.
Under ERISA (covering ~60% of insured Americans via employer-sponsored plans), the administrative appeals record is the only record in any subsequent federal lawsuit. If an automated appeal misses evidence, misframes the argument, or makes errors, those may be permanently locked in. An ERISA attorney explicitly warns: "Submitting an AI-drafted appeal without review can damage your case and might permanently lock in errors." This is the strongest liability argument against full auto-submission.
Medical billing advocates commonly charge contingency fees for insurance disputes. However, when a non-attorney takes a percentage of a legal recovery — especially in ERISA or bad-faith insurance contexts — this begins to resemble attorney fee-sharing, which is prohibited in all 50 states (ABA Model Rule 5.4). NYSBA Ethics Opinion 1271 (Aug 2024) specifically affirmed that paying a % of legal fees to a non-lawyer online service violates Rule 5.4(a). The key question: are we recovering an administrative denial (probably OK) or a legal claim (much higher risk)?
The ABA's first formal AI ethics guidance covers attorneys using generative AI — requiring competence, confidentiality, and human supervision. It doesn't directly govern non-lawyer AI document tools. No state has specifically prohibited AI-generated insurance appeal letters. The ABA's historical stance that non-lawyer document prep is UPL has been substantially softened by carve-outs in most states for standardized documents and form assistance.
Every existing competitor generates a letter for the patient to review and send themselves. UPHELD's model (auto-generate AND auto-submit via FHIR API) is meaningfully different. When the patient reviews and chooses to submit, UPHELD is clearly a document-preparation tool. When UPHELD submits autonomously, it looks more like active representation — which triggers UPL analysis more acutely.
Non-attorneys can act on behalf of a patient in administrative proceedings if the patient grants written authorization. ERISA regulations (29 CFR 2560.503-1) explicitly allow beneficiaries to designate authorized representatives. If UPHELD operates under explicit written patient authorization as a designated representative, this creates stronger legal ground — the company isn't practicing law, it's executing the patient's documented instructions.
| Company | Model | Fee | UPL Risk Management |
|---|---|---|---|
| Claimable | AI appeal after 25-min patient survey; patient reviews + sends | $39.95 flat/appeal | Flat fee (not contingency); patient-initiated submission; never called a law firm |
| Counterforce Health | AI + expert team; NIH-funded nonprofit | Free | Nonprofit status; grant-funded; no financial stake in outcome; patient submits |
| Fight Health Insurance | Chatbot + document upload; patient submits | Free | Free tool framing; no commercial UPL motive; patient submits themselves |
| Patient Advocate Foundation | Human advocates help with appeals | Free (nonprofit) | Explicitly notes UPL limits; stays in "advocacy" lane; 501(c)(3) |
| ResolveMedicalBills | Human billing advocates | Contingency (% of savings) | Explicitly warns advocates can't give "legal advice"; stays in admin advocacy lane |
| ERISA Recovery | Former insurance pros recover underpayments | Performance-based | No attorney services claimed; targets billing disputes not legal claims; works alongside lawyers |
Pattern: Every player either (a) keeps fees flat and non-contingent, or (b) operates as nonprofit/educational, or (c) explicitly disclaims legal advice and keeps humans in the loop. None auto-submits on the patient's behalf autonomously.
| Risk | Level | Mitigation |
|---|---|---|
| UPL claim for generating appeal letters | 🟡 Medium | Frame as document preparation / patient advocacy. Clear disclosures. Never use "lawyer" or "legal advice." |
| UPL claim for contingency fee model | 🔴 High | Get legal opinion before finalizing fee model. Consider flat-fee or subscription pricing. If contingency, structure carefully with attorney guidance. |
| UPL claim for auto-submission without patient review | 🔴 High | Insert mandatory patient review + approval step. Patient clicks "Submit" — UPHELD does not submit autonomously. |
| FTC consumer protection action (false advertising) | 🟡 Medium | Never use "lawyer," "legal advice," "law firm" in marketing. DoNotPay is the exact cautionary tale to avoid. |
| ERISA liability for errors that close the record | 🟡 Medium | Prompt patient to review before submission. Prominently disclose ERISA record-closure risk. E&O insurance. Consider attorney review option. |
| State-specific UPL prosecution | 🟢 Low | No state has targeted health insurance appeal tools. Existing companies operate all 50 states. Standard disclaimers sufficient. |
| Class action (consumer protection) | 🟢 Low-Medium | Don't guarantee outcomes. Be accurate about success rates. Strong ToS. Avoid "robot lawyer" positioning. |