Noah AI CAIHL draft report

Noah AI: CAIHL reassessment

September 8, 2026. Published AI-assisted draft.

Noah AI supports direct research questions, scope clarification, and editable evidence. These patient-accessible functions must be distinguished from inferred institutional uses. The evidence does not establish universal absence of recourse or dominant downstream effects.

Scope and agency posture

Publicly accessible professional research agent. Evaluate a patient's direct research use separately from hypothetical downstream industry effects. Preserve August 13 baseline 55 and its corrected prior-model history.

Posture: Mixed.

Axis: 55/100, retained provisionally. No numerical recalibration was performed.

Sources and documented findings

• The current homepage describes scope and constraint clarification, cited research, editable workflows and downloadable underlying data. Its use cases and testimonials emphasize life-science professionals. It lists GPT, Gemini, Claude and Grok and offers a support/feedback route. These establish product orientation, not the prevalence or consequences of industry use.

• The current privacy policy still states that sensitive information is not processed, names advertising/retargeting categories, describes general AI processing and an opt-out, and allows deletion or anonymization after retention needs end. It does not explain the specific health-query and downstream-model conditions well enough to resolve the prior gap.

• The current terms exclude uses subject to HIPAA and other named sector rules. That is a service restriction, not proof that an individual consumer's question is prohibited or that patient data is sent to advertisers.

Patient authority

Inference from the documented workflow: In direct use, patients can define research goals and constraints and take editable evidence away. Professional market positioning does not erase that authority. Patients affected by another organization's analysis are a different deployment and cannot be assigned a blanket lack of recourse without evidence.

Critical capacity

Editorial assessment: Citations and downloadable interim analysis can help users challenge framing and conclusions. The baseline's assertion that nothing supports patient action is too absolute: research can support a patient's chosen action without a dedicated appointment or appeal feature.

Informed control

Editorial assessment: The policy remains poorly specific about sensitive research inputs and model recipients. Distinguish documented advertising categories from an unsupported claim that health prompts are disclosed to them. Search absence cannot confirm that independent evidence does not exist.

Assessment and limits

Retain a mixed assessment but remove assertions about its largest real-world patient effect, universal lack of notice/recourse and confirmed absence of evidence. These exceed a public-source review. Center actual patient goal-setting and evidence export alongside unresolved conditions.

Confidence: Low to medium, primary documents only.

Remaining uncertainty: Direct patient experience, clinical boundaries inside the app, processor terms, research-input use and actual institutional deployments remain unknown.

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: 2026-08-13. Historical assessment. Earlier claims are not automatically reverified by this publication.