Page Menu

Reports, Remediation, and Reassessment

The assessment report documents what was evaluated, how the evaluation was conducted, what was found, and the limitations of the findings. The report is specific to the evaluated activity, system version, configuration, learner group, and intended use.

Assessment Report Contents

  • The assessment scope and intended use
  • The system version, configuration, workflow, and evaluation dates
  • The scenario package and assessment methods
  • Findings across the evaluation domains
  • Material strengths, limitations, and failure patterns
  • Clinically important errors, omissions, or evidence concerns
  • Any severity classifications or escalation findings used in the assessment
  • Considerations for remediation, monitoring, or reassessment
  • The limitations of the assessment and appropriate interpretation of the results

Controlled Factual Review

Before the report is finalized, the applicant may review a draft for factual accuracy concerning the system, version, configuration, workflow, and assessment conditions. The factual review does not change the predefined evaluation criteria or transfer control of clinical ratings and conclusions to the applicant.

Remediation

An applicant may address identified deficiencies and request additional testing. The AI Assurance Lab determines whether the changes can be evaluated through a limited reassessment or require a new full assessment. The scope, required materials, schedule, and fee for additional work are confirmed in writing before reassessment begins.

Changes That May Trigger Reassessment

  • A new model or system version
  • Material changes to system instructions, prompts, or response-generation logic
  • Changes to retrieval sources, citation functions, or evidence-update processes
  • Changes to guardrails, escalation rules, or safety controls
  • Changes to the intended use, learner population, clinical domain, or workflow
  • Changes that may affect commercial influence, privacy, logging, or responsible oversight

Important Limitations

  • The 25-scenario package is purposive and does not represent every possible learner interaction or clinical condition.
  • Findings apply only to the evaluated system version and configuration.
  • The assessment does not establish patient outcomes or clinical effectiveness.
  • Evidence verification is targeted to the evaluated responses and does not constitute a complete audit of every underlying source or database.
  • The assessment does not evaluate the entire company, model family, or product portfolio.
  • The assessment does not constitute certification, approval, accreditation, endorsement, a regulatory determination, or assurance of future performance.