Patient care navigation and health copilot AI
Ari
Ari offers rare-disease families practical care-navigation and advocacy assistance. The assessment is Mixed: control over tasks does not establish control over every contribution or derived output. Citizen Health and RARE-X data-governance arrangements require separate consideration.
AI-assisted public-source draft. Publication authorized by Hugo Campos. No comprehensive human verification or live product testing.
Summary judgment · Unscored in current comparisons
Qualitative draft. Deployment-specific consent and action controls remain unresolved.
Mixed
Potentially useful patient-directed assistance coexists with contractual limits on data authorship, prospective research withdrawal and unresolved release-specific approval controls. This review remains unscored.
Patient agency
How this tool changes agency
Could support understanding, record access and coverage appeals.
Control of the health narrative, effective correction and consent enforcement remain qualified or untested.
Patient agency assessment
Who sets and changes the goal?
Families choose care-navigation tasks. Rights in their contributions and derived information complicate control of the resulting health narrative.
Can patients tell AI is involved?
Ari is explicitly presented as an AI teammate. User comprehension of its limits was not tested.
Can patients meaningfully choose?
Published consent mechanisms support choice. Their application to legacy RARE-X data and current Ari enrollment remains unresolved.
Can patients correct or challenge the output?
Documented access and update rights do not establish reliable correction of an AI interpretation or its effect on subsequent actions.
Does it help patients understand or act?
Record interpretation and appeal preparation could support practical advocacy. Task completion, accuracy, citations and approval enforcement were not tested.
Can patients evaluate the conditions of use?
Public documents expose relevant conditions, but ownership categories, consent scope and withdrawal timelines need clearer deployment-specific explanation.
Text findings
What happens to patient data?
Different rights for different data
The report distinguishes medical records, self-reported information, derived outputs and research datasets. It examines licensing, access, deletion, retention and the limits of withdrawal. Legacy RARE-X agreements were not reviewed.
Who is left out or burdened?
Partially assessed
Digital access, caregiver capacity and premium-tier boundaries may affect benefit. Disability access and multilingual performance were not tested.
Are the clinical boundaries clear?
Partial
The stated role excludes diagnosis and prescribing. Approval before outward action is described differently in the partner FAQ and alpha terms. Application to the current release is unresolved.
Who defined what good looks like?
Not established for Ari outcomes
Community participation does not demonstrate patient authority over evaluation criteria or deployment decisions.
Review provenance
Criteria
CAIHL-derived HugoScore framework and current qualitative reassessment method
Reviewer
AI-assisted public-source review in OpenAI Codex
AI / model
OpenAI Codex / GPT-6
Human review
Hugo Campos authorized publication on September 17, 2026. This is publication approval, not comprehensive human verification.
Review date
2026-09-17
Limitations
No account creation, patient-data upload, live product testing, security audit, clinical evaluation or legal determination. Legacy RARE-X migration agreements and actual consent screens were not reviewed.
Review method
Public-source review of the product, terms, privacy policy, Databank consent and partnership announcement. Qualitative agency assessment with no numerical ranking.
Draft profile · Medium for documents, low for implementation (AI-assisted draft)