PHARMORIS ASSURANCE
Human authority and evidence boundaries remain visible.
This page explains the intended human, tenant, source, safety, data, release and future local boundaries around PHARMORIS. It describes target controls and open evidence; it does not claim certification, NHS approval, production readiness or completed assurance.
Target controls · Open evidence · No certification claimIDENTITY, TENANT AND ROLE
Server derived authority is a target control, not a completed claim.
The target model requires authenticated server policy to derive user, tenant and role context. A browser value, route label or representative role map is not authoritative. Complete isolation and permission behaviour must be proved for the exact candidate before it is called working.
Identity
Authenticate the person and preserve the actor and role at the action.
Target controlTenant
Enforce tenant scope on the server and prove cross tenant reads, writes, aggregates and exports fail.
Target controlRole
Authorise the exact action rather than trusting a client side role label.
Target controlFailure
Permission denied, expired and invalid session states must fail closed without exposing protected data.
Target controlSOURCE RIGHTS AND PROVENANCE
A definition or receipt is not an operational adapter.
Each used source needs a named owner, permitted purpose, receipt or approved snapshot, validation and quarantine route, freshness rule, tenant and export boundary and exact relationship to the evidence displayed.
If the source, permission or current evidence is not established, the truthful state is UNKNOWN.
Rights and purpose
Record who owns the source, how it may be used and where it may be displayed or exported.
Receipt or snapshot
Retain the exact version, retrieval time and integrity record needed to identify the material.
Validation and freshness
Keep invalid, stale or unauthorised material quarantined or unavailable.
Immutable link
Retain the relationship between the source basis, CandidateSignal and ActionCase.
CURRENT PRODUCT AND DATA BOUNDARY
Staff facing, bounded and non production.
The August 2026 scope excludes patient facing functionality, patient identifiable data and EPR write back. PHARMORIS is not presented as autonomous prescribing, patient specific advice or point of prescribing clinical decision support.
- 01
No patient login
The current product scope is staff facing.
- 02
No patient identifiable data
The present scope must not use patient identifiable information.
- 03
No EPR write back
Writing back to an electronic patient record is outside the August scope.
- 04
No autonomous decision
The product does not prescribe or replace an authorised person.
- 05
No deployment inference
A representative interface is not proof of a live customer or NHS environment.
EXACT HEAD RELEASE EVIDENCE
A candidate is no go while mandatory evidence remains open.
A non production release decision must identify one frozen candidate, its exact component and deployed identities, environment, tenant scope, schema, configuration, source register and evidence pack.
Any required blank, FAIL, MISSING, PENDING, BLOCKED or INVALIDATED evidence means no go. A screenshot, plan, mock, historic test or result from another version cannot satisfy the candidate.
Release identity
Reconcile repository, build, image, schema, configuration and running environment to one candidate.
Negative paths
Prove permission, tenant, failure, source, export, recovery and rollback behaviour at that same candidate.
Assurance review
Record the accountable reviewers, date, scope, residual risks and exact decision.
Invalidation
Revalidate when code, configuration, environment, intended use, evidence or scope changes.
RESPONSIBILITY MAP
Supplier controls and future local approvals must not be conflated.
NEUVIOR is responsible for evidencing the product, version, controls and boundaries it may supply. If an organisation evaluates or adopts PHARMORIS in future, that organisation would remain responsible for local intended use, information governance, clinical safety duties where applicable, procurement, configured testing, training, operational acceptance and any go live decision.
Supplier evidence
Freeze the intended purpose and exclusions; identify the exact product and environment; evidence controls, tests, source status, operations, support and known limitations.
NEUVIORLocal decision
Complete local data protection, information governance, clinical safety work where applicable, procurement, configured testing, training and go live approval.
Future local organisationEvaluation boundary
Agree intended use, authorised roles, evidence exchange, measures, escalation, stop conditions and language for a time limited evaluation.
Shared scopeCLAIM BOUNDARY
A control target or public record is not product certification.
No claim is made here of NHS approval, DTAC completion, DCB0129 completion, local DCB0160 completion, medical device status, generic certification, completed clinical safety approval or production deployment.
Intended use first
Clinical safety and regulatory applicability depend on the exact intended use and functions being assessed.
Product evidence
Organisation level records do not become evidence that a named product or version is approved.
Adopter evidence
Supplier evidence cannot grant a future local organisation's approval or go live decision.
Current wording
Open evidence remains open. A target, template or pending row is not a pass.
CURRENT EVIDENCE POSITION
The target is defined. The release proof remains open.
The supplied V5 evidence workbook records operational gates as pending and does not satisfy its no go formula. The exact head manifest is a control template, not a completed release record.
- 01
Exact candidate and running identities
Reconcile the same approved candidate across code, build, schema, configuration and environment.
- 02
Tenant, role and failure evidence
Complete the applicable negative paths for the frozen candidate.
- 03
Source by source evidence
Establish rights, receipts, validation, monitoring and operational behaviour for every used source.
- 04
Named decisions
Record accountable product, release, security, data, accessibility and other applicable reviewers.
- Podcast appearance. AI Ethics Now / University of Warwick IATL. AI and Accountability: When Does a Suggestion Become a Decision?. 23 August 2026.
- Contributed opinion. TechRadar Pro. Five questions to test whether an AI stack is truly under your control. 11 August 2026.
- Founder commentary. ET Pharma. India’s pharma quality reset will be won after the inspector leaves. 4 August 2026.
- NEUVIOR analysis. NEUVIOR Insights. From pilot theatre to financial grip: how NHS CFOs should evaluate AI in medicines optimisation. 18 March 2026.
- NEUVIOR analysis. NEUVIOR Insights. NHS digital programmes do not fail on dashboards. They fail on deployment. 20 March 2026.
- Expert commentary. Manufacturing Chemist. NICE approval is not enough: why the UK's next medicines challenge is operational delivery. 4 June 2026.
- Founder commentary. PharmaTimes. How we roll. 10 June 2026.
- Editorial roundup. Digital Health. Digital Health Coffee Time Briefing. 26 May 2026.
Classified public record
ANALYSIS + EVIDENCE BOUNDARY
Read the argument. Verify the product separately.
Published analysis can explain why governance matters. It cannot evidence security, clinical authority, certification, product readiness or local approval. Use the dated NEUVIOR records for that.
AI and Accountability: When Does a Suggestion Become a Decision?
Varun Sharma joins AI Ethics Now to examine false confidence, human responsibility and the point where AI assisted suggestions become accountable decisions.
Hear where an AI suggestion becomes a decisionFive questions to test whether an AI stack is truly under your control
In TechRadar Pro, Varun Sharma sets out five tests for enterprise AI control across provenance, infrastructure, legal authority and safe exit.
Test provenance, control and safe exitFrom pilot theatre to financial grip: how NHS CFOs should evaluate AI in medicines optimisation
NHS finance leaders do not need another AI pitch. They need a disciplined way to decide whether a technology can release value within the financial rules they are already being held to.
See how a finance leader should test an AI claim
NEUVIOR Insights and After Publication reflections are company or founder analysis. Publisher records evidence publication, not endorsement, adoption, product readiness, savings or clinical outcomes.
A BOUNDED REVIEW
Start with the intended use, exact version and decision.
A useful assurance discussion identifies the product version, environment, intended use, data scope, authorised roles and decision the evidence is meant to support. Do not submit patient identifiable, confidential or security sensitive material through a public enquiry.