HugoScore hugoscore.org

Full review

My Doctor Friend CAIHL draft report

Evidence-linked HugoScore draft report for a health AI tool that affects patients.

Agency checklist · October 30, 2026

Leans expands 9 of 12 yes

Document-based draft. Independent confirmation is incomplete. Hugo has not personally verified this service.

Assessed function: Patient and caregiver companion. Assessed for self-started consumer use. Provider-mediated use and community sharing noted as separate contexts.

Initial Anthropic coding has an Anthropic model in the role record. Astra review conflicted: OpenAI model in role record.

A no can mean missing documentation or institution-dependent control; it does not by itself establish harmful behavior. These categories are not clinical safety or product-quality ratings.

  • Public documents do not establish actual product behavior.
See all twelve answers and their evidence
QuestionAnswerEvidence / reviewSource and reasoning
1. Told it is AIYesdocumented
Confirmed at document level
may include answers generated by artificial-intelligence models Source

Terms section 2. Homepage also says three AI models answer each message.

2. Operator and model maker namedYesdocumented
Confirmed at document level
AI processing (OpenAI, Anthropic, and Google) under HIPAA Business Associate Agreements. Source

Operator My Doctor Companion, Inc. is named. Homepage says which models are used 'depends on your question, your membership plan, and other factors' (FAQ).

3. Patient chooses to use itYesdocumented
Confirmed at document level
It’s a tool your patients hold, not a dashboard you manage. Source

Assessed for self-started patient or caregiver use. Clinicians may recommend it but the patient takes it from there. Provider-mediated (BAA) use is a separate context not assessed here.

4. Patient sets the taskYesdocumented
Confirmed at document level
doctor visit preparation, answering general health questions, test result interpretation Source

Patient chooses the use. The product asks follow-up questions as a safety filter.

5. Patient can decline or stopYesdocumented
Confirmed at document level
It’s a tool your patients hold, not a dashboard you manage. Source

Self-started: Q3 is yes and no care depends on using it. Source is the Q3 source. Terms 3 allow deleting the account any time.

6. Patient sees the AI outputYesdocumented
Confirmed at document level
You leave with a full transcript, a structured summary Source

Chat answers are shown directly to the user. Audio is not stored.

7. Patient can correct or challengeYesdocumented
Confirmed at document level
Access, correction, or portability – email [email protected] Source

Generic email route to the privacy contact, plus the homepage says users can fix record errors once in the app. No output-specific dispute mechanism documented. Reviewer judgment call.

8. Patient can get a copyYesdocumented
Confirmed at document level
Access, correction, or portability – email [email protected] Source

Clinicians page also mentions exporting a PDF and a quick link. Terms 4 restrict bulk exporting outside the official API.

9. Patient can deleteYesdocumented
Confirmed at document level
Delete Account – in the app: Settings > Account > Delete Source

Primary deletion within 30 minutes, AI model logs within 24 hours, backups purged on next rotation. Moderation audit log may be kept.

10. Training or sale does not override patient choiceNonot-documented
Corrected after independent review; accepted by editor
our providers are contractually barred from training their own models on it Source

The restriction covers model providers. An operator no-training promise or opt-out remains undocumented, not proven absent in practice.

11. Patient can run it or take it elsewhereNonot-documented
Confirmed at document level
Quotation not recorded; evidence repair pending. Source

Vendor-hosted app. Export or PDF only. No self-hosting or documented move-elsewhere route.

12. No advertising or data businessNodocumented
Confirmed at document level
use or share that data for analytics, product development, or research Source

Privacy Notice states it never sells personal information or shares it for cross-app advertising, and the clinicians page says it never uses patient data for advertising. But de-identified sharing for research is documented and the notice does not say it is not for revenue. Conservative no. Reviewer may flip.

Initial coding: Claude Sonnet helpers, compiled by Claude Opus 5.5 (Anthropic). Selective review and corrections: GPT-6 Astra (OpenAI). Editor: Hugo Campos. Acceptance of corrections is not product verification.

Checklist rules · Independent review and limitations · Later evidence follow-up

Earlier report

My Doctor Friend: CAIHL reassessment

September 8, 2026. Published AI-assisted draft.

My Doctor Friend combines patient questions, records, and visit support. The current privacy notice distinguishes consumer and provider use, names Google processing, and describes derivative-data permissions and layered retention. The earlier unread-policy caveat is replaced by these conditions.

Scope and agency posture

Patient and caregiver companion, with provider-mediated use and community sharing as distinct contexts. Preserve the August 9 baseline 82. This pass resolves the prior unread-privacy-notice gap.

Posture: Potentially agency-expanding.

Axis: 82/100, retained provisionally. No numerical recalibration was performed.

Sources and documented findings

  • Public web retrieval returned empty bodies. One targeted browser rendering successfully read the June 20 privacy notice, without login or interaction with the product.
  • The notice distinguishes direct consumer use from provider business-associate processing. It names OpenAI, Anthropic and Google, with stated BAA arrangements, and permits non-personal derivatives for analytics, development or research. It documents account deletion, a correction/portability contact and analytics opt-out.
  • Account/health-record retention includes up to seven years in active storage and six-year encrypted backup language. Deletion promises distinguish primary data, AI logs and backup rotation. Community contributions are visible to members under pseudonyms. Its Apple Health noncollection statement is explicitly a launch-era statement, leaving the current integration unresolved.

Patient authority

Inference from the documented workflow: Patient-selected conversation and preparation remain the main agency basis. Consolidation alone does not prove patient custody independent of the vendor. Caregiver control also needs to be distinguished from the authority of the person whose information is held.

Critical capacity

Editorial assessment: The baseline's preparation and thought-partner rationale is plausible. Data correction rights now have a current source, but a right to request correction is not proof of editable generated summaries or reliable error resolution.

Informed control

Editorial assessment: The privacy gap is now narrowed substantially. Direct-use safeguards should not be described as universal HIPAA coverage, and no-sale language should not imply no derivative research use. Retention and community visibility deserve plain-language prominence.

Assessment and limits

Change the stale unread-notice finding into a substantive description of the published conditions. Keep the qualified patient-directed posture, without treating founder identity or record consolidation as proof of independent control.

Confidence: Medium for rendered policy, limited for product behavior.

Remaining uncertainty: Current audio and Apple Health processing, backup rotation duration, training restrictions and summary correction remain unresolved. Pricing and App Store counts remain August evidence.

Published documentation is evidence of stated conditions, not proof of actual implementation. Unknowns did not receive automatic negative points. Funding, sponsorship, and public code do not determine agency by themselves.

Review provenance

  • Reviewer/model: OpenAI Codex / GPT-6.
  • Method: Focused public-source reassessment using CAIHL: patient authority, critical capacity, and informed control. Existing evidence plus one focused primary-source pass and at most one targeted follow-up. No live product testing. Numerical scores remain provisional editorial placements, not a new calculation.
  • Human review: Hugo Campos authorized publication of these AI-assisted draft reassessments on September 8, 2026. This does not claim comprehensive human verification of every finding.
  • Earlier review: 2026-08-09. Historical assessment. Earlier claims are not automatically reverified by this publication.