CureWise CAIHL draft report

HugoScore CAIHL Draft Report: CureWise

• Status: Draft for human review
• Last reviewed: 2026-07-10
• Review method: Public-source review of CureWise product, pricing, privacy, terms, press, and public onboarding surfaces; review of interview and launch coverage; comparison of trial-count claims with ClinicalTrials.gov; and review of FDA and FTC context. No paid account, record connection, authenticated workflow test, model/data-flow inspection, vendor interview, patient interview, oncologist review, security audit, accessibility audit, independent validation, or legal/regulatory determination was performed.
• Service: CureWise
• Vendor: CureWise, Inc.
• URL: https://curewise.com/
• Category: Condition-specific coaching and care AI

1. Executive Summary

CureWise is a paid, direct-to-patient AI cancer navigator publicly launched in June 2026. It says users can connect records from U.S. health systems and labs, combine those documents into a cancer timeline, ask questions in plain or technical language, compare answers from Claude, Gemini, and GPT, identify personalized treatment possibilities, match with clinical trials, and prepare questions for clinicians.

Under CAIHL, CureWise is strongly patient-directed: the patient chooses the tool, brings their own records and questions, and uses the output to understand and advocate. That can support critical reflection and strategic action. The service is vendor-hosted, however, and processes full medical and genomic information through a data flow that is not fully public. The privacy policy does not identify the external AI providers or explain what patient context they receive, their retention terms, or model-training restrictions.

Clinical boundaries are also mixed. CureWise repeatedly says the service is educational, can make mistakes, and should not replace medical advice. Yet its how-it-works page shows a patient-specific question about restarting a cancer drug, followed by concrete suggestions about timing, supportive medication, laboratory monitoring, and supplements. No CureWise-specific peer-reviewed validation, independent matching-accuracy study, or regulatory review was identified in this pass.

• Agency posture: Potentially agency-expanding, mixed
• Agency axis position: 84 of 100
• Confidence: Low-to-medium draft, official sources plus interview and launch coverage

2. CAIHL Question

Who does CureWise serve?

CureWise primarily serves adult cancer patients and authorized care partners who choose and pay for the service. It also serves CureWise's subscription business and depends on external record-connectivity and general-purpose AI providers. Clinicians are not the primary users, but patient-shared questions and outputs may influence clinical conversations and treatment or trial decisions.

CAIHL classification: Patient-directed, vendor-hosted cancer navigation AI.

3. What The Service Does

Public materials describe a workflow that begins with medical-record connection or document upload. CureWise says it synchronizes records in the background, creates a longitudinal cancer story, explains results at the user's preferred level of detail, and answers questions using the user's documents.

Cross-Check submits one question to Claude, Gemini, and GPT, then shows and synthesizes the responses. Trial matching uses cancer type, stage, mutations, and treatment history, with stated match reasons and eligibility explanations. The public price on July 10, 2026 was $49 per month or $439 per year after a seven-day trial. Accounts are limited to adults.

4. Patient-Impact Pathway

1. An adult patient or authorized representative chooses CureWise and begins subscription onboarding. 2. The user provides account, profile, payment, health, genomic/genetic, and cancer-related information. 3. The user connects health systems or labs and/or uploads documents. 4. CureWise consolidates the record into a timeline and case summary. 5. The system explains results, answers questions, compares multiple model outputs, and matches trials. 6. The patient may share outputs or use them to prepare for oncology conversations. 7. Agency depends on accuracy, current evidence, correction routes, clinical review, and meaningful control over every data recipient.

5. Evidence Table

| Source | Evidence | CAIHL relevance |
| --- | --- | --- |
| CureWise homepage, accessed 2026-07-10: https://curewise.com/ | Describes an AI cancer navigator combining health documents, research, and trials into personalized guidance. | Supports product identity and patient-agency purpose. |
| How it works, accessed 2026-07-10: https://curewise.com/how-it-works | Describes record syncing, cancer timelines, adjustable explanations, document-grounded answers, Claude/Gemini/GPT Cross-Check, and trial matching. | Core product, visibility, control, action-support, and clinical-boundary evidence. |
| Pricing, accessed 2026-07-10: https://curewise.com/pricing | Lists a seven-day trial, $49 monthly plan, and $439 annual plan. | Supports choice and equity findings. |
| Privacy policy, last updated June 3, 2026: https://curewise.com/privacy | Discloses broad health, genomic, payment, usage, and log-data collection; user rights; service-provider sharing; non-sale; and claimed security controls. Says CureWise is not a HIPAA covered entity. | Core data-governance and control evidence. |
| Terms, last updated June 3, 2026: https://curewise.com/terms | Says outputs may be inaccurate or outdated, requires clinician confirmation, preserves user ownership of health information, and permits some de-identified/aggregated retention after deletion. | Supports clinical limits, ownership, residual-data, and reliance-risk findings. |
| Public onboarding, accessed 2026-07-10: https://app.curewise.com/onboarding | Shows an active eight-step patient onboarding flow and seven-day trial. No information was submitted. | Confirms direct patient access and review limits. |
| CureWise launch release, June 22, 2026: https://www.globenewswire.com/news-release/2026/06/22/3315157/0/en/curewise-founder-and-ceo-steve-brown-to-speak-at-aspen-ideas-health-2026-and-launch-ai-powered-precision-oncology-platform.html | Company-issued release identifies the public launch and multi-model approach. | Supports launch timing; not independent validation. |
| CURE interview, November 18, 2025: https://www.curetoday.com/view/ai-tools-help-patients-navigate-complex-cancer-information | Describes founder and patient experiences with cancer-navigation AI. | External editorial context and narratives, not controlled validation. |
| The ASCO Post interview, January 25, 2026: https://ascopost.com/issues/january-25-2026/using-ai-to-ensure-that-all-patients-with-cancer-have-access-to-precision-oncology-care/ | Founder describes intended use for understanding disease, forming questions, and finding trials. | Supports intended use; claims remain self-reported. |
| ClinicalTrials.gov trends, July 2, 2026 data: https://clinicaltrials.gov/about-site/trends-charts | Lists 592,210 total registered studies and 64,732 recruiting studies. | CureWise's “500,000+ active” wording needs clarification. |
| FDA CDS overview: https://www.fda.gov/media/161775/download | Summarizes the health-care-professional focus and other requirements of the non-device CDS pathway. | Regulatory context only; no legal conclusion about CureWise. |
| FTC health-app guidance: https://www.ftc.gov/business-guidance/resources/complying-ftcs-health-breach-notification-rule-0 | Explains privacy duties that may apply to non-HIPAA personal-health-record vendors. | Data-governance context only; no legal conclusion about CureWise. |

6. Mixed HugoScore Profile

Who does this AI serve?

Patient-directed, vendor-controlled. Adult patients choose and pay for CureWise, and no institutional or life-sciences sponsor pathway is disclosed. Vendor and external model-provider interests remain part of the control structure.

Can patients tell AI is involved?

Yes. The AI role, model names, multi-model comparison, AI-generated output, and error warnings are explicit.

Can patients meaningfully choose?

Partial to yes. Use is voluntary, users may cancel, and public materials say connected-system access can be revoked. The subscription price and unclear downstream model controls limit practical choice.

Can patients correct or challenge what the AI produces?

Partial. The privacy policy offers personal-data access, correction, deletion, portability, and objection rights. Public materials do not show how a user edits a generated timeline, corrects a biomarker or treatment-history extraction, disputes a treatment analysis, or challenges a trial match.

Does it help patients understand or act?

Yes, with high-stakes caveats. Plain-language explanations, longitudinal records, appointment questions, multi-model comparison, and trial discovery can support patient action. CureWise-specific accuracy and outcome evidence are not public.

Who is left out or burdened?

Full use costs $49 per month or $439 per year, record connection is U.S.-focused, and accounts are limited to adults. Public sources do not establish multilingual support, financial assistance, accessibility-standard conformance, low-literacy testing, or robust caregiver/proxy controls.

What happens to patient data?

CureWise describes broad collection of records, labs, genomic/genetic data, images, messages, account and payment information, and usage data. It states that data is not sold, sensitive health data is not used for third-party advertising, and users have access, correction, deletion, restriction, objection, and portability rights. It also claims encryption, role-based access, and audit logging.

Material gaps remain. The privacy policy does not name the AI model providers or explain what Cross-Check sends to them, model-provider retention or training restrictions, named record-connectivity and cloud subprocessors, or specific retention periods. Terms allow residual de-identified or aggregated derivatives after deletion.

Are the clinical boundaries clear?

Clear in legal text, blurred in product examples. CureWise warns that outputs may be wrong and are not medical advice. The public medication-restart example nonetheless gives concrete patient-specific suggestions that could shape treatment behavior.

Who defined what good looks like?

Mostly CureWise and its founder. Public evidence includes company materials, launch coverage, interviews, testimonials, and one clinician testimonial. No CureWise-specific peer-reviewed performance, patient-safety, equity, or patient-agency evaluation was identified.

7. Key Unknowns

• The vendor and governance behind the “70,000+” record-connection claim.
• The definition and status filters behind “500,000+ active clinical trials.”
• Trial-match accuracy, update frequency, false-negative handling, and clinician review.
• Whether patients can edit record-derived timelines and extracted clinical facts.
• What patient context is sent to Anthropic, Google, and OpenAI.
• External-model retention, training, logging, location, and contractual safeguards.
• Named subprocessors and specific retention/deletion periods.
• Clinical safety testing and escalation for conflicting or high-risk outputs.
• Independent security, accessibility, outcome, and patient-agency evaluation.
• Qualified regulatory assessment of patient-facing treatment-option and matching functions.

8. Patient Agency Interpretation

CureWise begins on the patient-directed side of the CAIHL axis: patients choose it, bring their own records, and use it to formulate their own questions. A reliable version could help a patient turn an opaque oncology record into a coherent narrative and then use that narrative to seek clarification, a second opinion, a trial, or a correction in care.

The central risk is that apparent clarity may exceed warranted confidence. Agreement among three general-purpose models is not independent clinical confirmation. For high-stakes cancer questions, meaningful agency requires source traceability, visible uncertainty, editable records, challenge routes, granular data controls, and clinician escalation—not only a disclaimer.

9. Publication Recommendation

Ready for human review as an AI-assisted, source-backed draft. Publish under Condition-specific coaching and care AI with CAIHL classification “Patient-directed, vendor-hosted cancer navigation AI.” Do not mark reviewed or verified. Prioritize human review of the trial-count wording, medication-guidance boundary, model-provider data flows, output correction and challenge, clinical validation, accessibility, security evidence, and regulatory posture.

Review Provenance

• Criteria: HugoScore patient agency framework derived from CAIHL, using the same public questions and mixed answer types applied to every tool.
• Reviewer: AI-assisted public-source draft prepared in OpenAI Codex; no named human reviewer is recorded.
• AI / model: OpenAI Codex / GPT-5.
• Human review: No comprehensive human review has been completed or claimed.
• Review date: 2026-07-10.
• Limitations: No paid account, record connection, authenticated workflow test, vendor or user interview, clinical review, security audit, model/data-flow inspection, accessibility testing, independent validation, or legal/regulatory determination.