⚖ JSP 936 V1.1 — Dependable AI in Defence

NOVA™ Compliance
Documentation

Full supplier-level compliance documentation for MOD procurement teams, TLB RAISOs, and defence prime contractors. Prepared in accordance with JSP 936 V1.1 (November 2024) and the Ambitious, Safe, Responsible (ASR) Policy 2022.

Responsible Person Will Kennedy-Long FLPI, FITOL
Overall Risk Rating MODERATE — TLB-Level Oversight
Classification Scope OFFICIAL Only
Document Version v1.1 — 18 October 2026
Next Review March 2027
⛔
CLASSIFICATION BOUNDARY — OFFICIAL ONLY

NOVA™ is for OFFICIAL material only. It runs on Cloudflare, with Anthropic, Voyage AI and Pinecone as AI and search processors. It must not be used to process OFFICIAL-SENSITIVE, SECRET, or TOP SECRET material under any circumstances. This is a hard operational constraint and must be communicated to all users within your organisation.

ASR 2022 — Five Ethical Principles
Ethical Principles Assessment
Residual risk ratings per principle following mitigation. Full detail in NOVA-AERA-001.
✅
Human-Centricity
Minor
✅
Responsibility
Minor
⚠️
Understanding
Moderate
✅
Bias & Harm
Minor
⚠️
Reliability
Moderate
Supplier Documentation Pack
Download Compliance Documents
All documents prepared per JSP 936 V1.1. Available for MOD procurement review.
NOVA-AERA-001
AI Ethical Risk Assessment
Full assessment against all five MOD AI Ethical Principles. Stakeholder impact analysis, risk ratings per principle, mitigations register, and supplier assessment of MODERATE (TLB-Level) risk per JSP 936 Table 1. Final determination rests with the customer RAISO.
JSP 936 §86ASR 2022Five PrinciplesRisk Register
✓ Complete ⬇ Download DOCX
NOVA-MC-001
Model Card
Documents every AI component in the code (Claude Sonnet 4.6 and Claude Haiku 4.5, Voyage voyage-large-2 with nine Pinecone indexes, and the Cloudflare Workers AI picture and speech models), the Operational Design Domain, known limitations, and data handling per component. Version 1.1, 18 October 2026. Required MVP per JSP 936 Cover Note (Nov 2024).
JSP 936 Cover Note§63 UnderstandingAI ComponentsODD
✓ Complete ⬇ Download DOCX
NOVA-SAEA-001
Statement of AI Ethics Assurance
Annual supplier-level statement supporting any MOD TLB Executive Board's Statement of AI Ethical Assurance to 2PUS. Tick-by-tick confirmation against every JSP 936 requirement with evidence references. Valid until March 2027.
JSP 936 §99TLB SupportAnnual StatementRAISO
✓ Complete ⬇ Download DOCX
NOVA-BHMA-001
Bias and Harm Mitigation Analysis
Analysis of LLM training data bias, RAG retrieval bias, and five harm categories. Includes Anthropic Constitutional AI mitigations and a monitoring schedule with quarterly doctrine index reviews.
JSP 936 §69–70Bias AnalysisHarm MitigationRAG Architecture
✓ Complete ⬇ Download DOCX
NOVA-SRA-001
Security Risk Assessment
Risk register covering the data flows to every processor: Cloudflare (including Workers AI), Anthropic, Voyage AI, Pinecone, Resend and Mailgun, and GitHub. Includes the classification boundary, the handling of uploaded document text as untrusted data in prompts, UK GDPR incident response, and guidance for MOD IT security teams. Version 1.1, 18 October 2026.
JSP 936 §195–197Secure by DesignData FlowsOFFICIAL Only
✓ Complete ⬇ Download DOCX
NOVA-UTP-001
User Training Package
Operator training material covering NOVA system behaviour, performance, known limitations, and trust calibration. Required per JSP 936 §130. Not yet available: the target of Q2 2026 was missed.
JSP 936 §130User TrainingTrust Calibration
⚠ In Development ⬇ Not yet available
Full Requirements Mapping
JSP 936 V1.1 Compliance Status
Requirement-by-requirement mapping against JSP 936 V1.1 (November 2024).
RequirementJSP 936 RefStatusEvidence
Human control embedded at every lifecycle stage§56–58✅ MetNOVA platform — Mark Complete mechanism
AI Disclosure on all generated documents§40✅ MetCover note + footer on all DOCX exports
AI Ethical Risk Assessment completed§86✅ MetNOVA-AERA-001
Model Card publishedCover Note✅ MetNOVA-MC-001
Statement of AI Ethics Assurance§99✅ MetNOVA-SAEA-001
Bias and Harm Mitigation Analysis§69–70✅ MetNOVA-BHMA-001
Security Risk Assessment (Secure by Design)§195–197✅ MetNOVA-SRA-001
Doctrine traceability in all outputs§63✅ MetCriteria system with JSP/DTSM clause refs
Defined Operational Design Domain§75✅ MetNOVA-MC-001
Classification constraint clearly stated§168✅ MetPlatform-wide + all documentation
User Training Package for AI system operators§130⚠️ PendingNOVA-UTP-001 — not yet available

Procurement & RAISO Enquiries

For MOD procurement teams, TLB RAISOs, or defence prime contractors requiring additional assurance documentation, technical briefings, or supplier due diligence — contact the responsible person directly.

✉ Contact Will Kennedy-Long