Disclosure presentation
Was a recorded disclosure presented before the specific AI action it references? This does not prove comprehension or legal consent.
Check disclosure evidence →
HEALTHCARE AI / CHECKABLE EVIDENCE
Care moves forward.
Keep the decision
context in view.
When clinical information changes, the original record still matters. Explore signed evidence components for what was recorded, transferred and changed around an AI workflow.
Public signed examples. No patient information. The clinical correction sequence is illustrative, not a validated clinical deployment.
CLINICAL SYSTEMS / A SEPARATE EVIDENCE RECORD
THE ORIGINAL CONTEXTWHAT CHANGED LATER
PreservedModel and source references
LinkedCorrection and review context
Proposed workflow. Missing evidence stays unknown. Care decisions stay with the care team.
Disclosure evidenceNotice and action, in recorded order
Custody evidenceTwo keys, matching transfer terms
Model-change evidenceEvaluated and reported serving stateWORKING COMPONENTS / NARROW CLAIMS
Choose the question.
Inspect the receipt.
Six receipt types, each with a defined boundary. The live checks below expose the signature, linking and ordering gates rather than a generic “trust” score.
Custody presentation + handoff
One party signs proposed terms. A handoff checks two distinct signing keys and matching terms. Keys alone do not establish organizational authority.
Check the two-party handoff →
Retention policy + purge
Check the recorded policy and its linked purge event. A receipt is not proof that every possible copy was erased.
Inspect the retention checks →Model change
Compare the evaluated configuration with the reported serving configuration. A match does not establish model safety or clinical accuracy.
Check the configuration record →LIVE VERIFICATION / SIGNED PUBLIC EXAMPLES
Don't take the label on trust.
Run the check.
Choose a receipt. The page submits only the supplied public example to Hive's production verifier and displays its returned gate-level result.
HIVE PRODUCTION VERIFIERREADY
Select a receipt to run the proof.
No patient information or upload is involved. This page demonstrates online verification by Hive, not an independently operated clinical system. A valid result establishes only the named receipt checks.
COMPLEMENTARY CAPABILITIES
Validate the model.
Preserve the interaction.
These are different questions, not competing answers. Evidence infrastructure complements clinical validation and governance; it does not replace them.
MODEL ASSURANCE
How does this AI perform?
Epic's AI Trust and Assurance Suite supports evaluation of performance, fairness, drift and clinical impact. Read Epic's validation guidance.
HIVE EVIDENCE
What does the record establish?
Check signed disclosures, handoff terms, retention records and model-change claims. Explore how original context and later corrections could remain linked for review.
Hive is not affiliated with or endorsed by Epic or UCSF Health. This public healthcare page does not represent a customer relationship or accepted clinical evaluation.
OBSERVATIONAL SIDECAR / FAIL-OPEN
Care continues.
Hive stays beside it.
In a proposed integration, existing clinical systems remain in control. If evidence collection is unavailable, care continues and the evidence gap remains explicit. Data scope, privacy and access controls must be agreed before any integration.
Clinical workflowNo dependency on HiveHive sidecarScoped evidence copyProposed architecture, not a live clinical connection.
FREE / 30 DAYS / SYNTHETIC DATA FIRST
One workflow.
A useful question to test.
Choose one corrected-information scenario and agree success criteria with your team. Evaluate whether the original context, correction and remaining unknowns are easier to review.
Based in Walnut Creek · available for a local working session
Evidence is not a clinical verdict. Signatures do not establish source truth, complete capture, clinical validity, causation, patient comprehension or regulatory compliance. No patient data is requested on this page.