AI Security in Production — Part 2: Supply Chain, Guardrails, Red Team & the Defense Program — 42 Templates, Checklists & Worksheets for the AI-BOM, Guardrail Control System, Red-Team Program, Detection & Incident Runbook, Compliance Crosswalk & Defense P

Trust no artifact, prove it, then survive it. 42 production templates, checklists & worksheets distilled from Chapters 7–12 of AI Security in Production plus its Appendix A checklist and Appendix B crosswalk: the AI-BOM and provenance gates, guardrails engineered as a control system, the six-suite red-team program with ASR gates, decision-record telemetry, SEV 1–4 classification and the gateway incident runbook, the book's own NIST / ISO 42001 / OWASP / EU AI Act crosswalk, and the defense program — owner, ninety days, metrics, board pack. Standalone: complete and useful without Part 1. PDF + Markdown source.

Your AI estate is assembled from other people's artifacts. Every one arrives pre-trusted.

Hosted models you cannot inspect, open weights you did not train, datasets you did not collect, plug-ins a stranger published last month — your deployment pipeline trusts all of them, which makes the supply chain the highest-leverage attack surface in the program and, in most enterprises, the least governed. AI Security in Production — Part 2 is the second half of the discipline, as a working pack: 42 copy-paste templates, checklists, and worksheets distilled from Chapters 7–12 of the dedicated security book, plus its Appendix A enterprise checklist and Appendix B quick-reference crosswalk — inventory and verify every artifact before it serves traffic, engineer guardrails as a control system with sensors, decisions, enforcement points, and logs, attack your own systems first under written rules with scored gates, assume breach and build the telemetry that reconstructs any conversation in five minutes, map the one control set onto the four frameworks your auditors speak, and run it all as a program with an owner, a ninety-day plan, and proof a board can read.

Every template carries the book's tested positions built in: an AI-BOM generated from the platform itself and reconciled quarterly, because drift between declared and observed is a finding; a behavioural fingerprint suite that notices when a hosted model changes underneath you; guardrails audited limb by limb, because a guardrail without an enforcement point and a log is a wish; red-team gates with the book's example numbers (extraction attack success rate at or below 2% for cross-tenant goals, injection-to-action at or below 1% for tool-using agents — the book's example gates, labeled as the book's, for you to set by your own risk appetite); severity written in outcomes a duty manager can apply at 3 a.m.; the book's own crosswalk table, carried in full at exactly the granularity the book prints it; and a board promise that never says "secure" — bounded, tested, and priced, with the evidence on one page. Where the book gives an example — its gates, its metric targets, its severity definitions — the template labels it as the book's example for you to refill with your own measurements. Bracketed fields are yours to fill; nothing here invents a threshold for your estate. This pack stands alone: it is complete and useful without Part 1, and nothing in it sends you elsewhere for an answer.

What's inside — 42 templates, checklists & worksheets in 7 sections

1. Trust No Artifact — Model Supply Chain, Provenance & Poisoning (6) — AI-BOM Inventory Template · Provenance & Integrity Verification Checklist · Four-Door Poisoning Defense Sheet · Dataset Datasheet, Licence & Documentation Gate Sheet · Vendor Assessment Scorecard — Six Questions, Tiering & Fallbacks · Artifact Lifecycle Sheet — Evaluation Gate, Plug-ins, Response, Retirement

2. Guardrails That Hold — Input, Output & Policy Enforcement as a Control System (6) — Four-Limb Guardrail Audit — Sensor, Decision, Actuator, Log · Input Guard Stack Specification — Beyond Injection · Hardened Generation Checklist — Prompts, Structure, Decoding · Output Verification Stack — Scan, Ground, Render · Policy Engine & Conflict-Precedence Sheet · Guard Operations Pack — Testing, Budgets, Streaming, Refusals

3. Attack Yourself First — The Enterprise AI Red-Team Program (6) — Rules of Engagement Template — Authorised, Scoped, Safe · Six Attack Suites Corpus Specification · Scoring & Release-Gate Sheet — ASR, Severity, and the Gate · Human Tester's Playbook & Pre-Cycle Intelligence Pack · Findings-to-Fixes Tracker — SLAs, Owners, and No Quiet "Won't Fix" · Corpus Factory, Extraction-Suite Design & Vendor-AI Testing Sheet

4. Assume Breach — Monitoring, Detection & the AI Incident Runbook (7) — Telemetry Standard — the Decision Record Specification · Detection Catalogue — the Six Starter Families · AI Incident Severity Classification Sheet — SEV 1 to SEV 4 (the book's severity table in full) · AI Incident Runbook — the First Sixty Minutes and Beyond · Tabletops & Conversational Forensics Sheet · Communications Playbook & Insider-Variant Response Sheet · Detection Engineering & Recovery Verification Sheet

5. The Compliance Map — NIST AI RMF, ISO 42001, OWASP & the EU AI Act (6) — Four Frameworks Orientation Sheet — Honestly Introduced · Control-to-Clause Crosswalk — the Book's Table 3, in Full · EU AI Act Classification Worksheet · Gap Analysis & Audit-Readiness Sheet — Compliance as Exhaust · Procurement Questionnaire & Certification-Sequencing Sheet · Living Crosswalk, Assurance Ladder & Auditor-Interview Prep Sheet

6. The Defense Program — People, Roadmap & Proof (6) — Program Governance Charter — Ownership With Teeth · First-Ninety-Days Plan — Stabilise, Systematise, Scale · Staffing & Budget Sheet — Honest Arithmetic · Metrics Pack — Proof Without Theatre (the book's metrics table, example targets labeled) · Board Conversation Pack — Risk, Trajectory, Decisions · Annual Operating Calendar, Culture & Continuity Sheet

7. Part 2 Gates & Working Artefacts — Appendices A & B (5) — Appendix A Audit Sheet — the Enterprise Checklist, Appendix Form (all 25 lines) · Appendix B Quick Reference — Crosswalk, OWASP List & Act Tiers · Part 2 Production Security Gate — Chapters 7–12 Sign-Off (37 lines) · Guardrail Change Log & Deprecation Review Sheet · Near-Miss Register & Ninety-Day Incident Tail Sheet

Who this is for

CISOs and security leaders building the second half of an AI defense program, security architects and platform leads who own guardrails, telemetry, and release gates, red-team and detection engineers, and GRC leaders who must answer auditors, regulators, and enterprise customers from evidence. The source book assumes you know how enterprise systems are generally built and defended; no machine-learning expertise is required. If you cannot say where your flagship model came from, whether it changed last week, and what would happen in the first sixty minutes after it misbehaves — start at Template 1.

What this pack is not

Not Part 1, and not dependent on it. Threat modeling, prompt-injection defense-in-depth, data-as-payload protection, identity and access for AI workloads, and agent tool security are the book's Chapters 1–6 and live in the Part 1 companion pack. Nothing here restates them, and this pack is complete and useful without them. Where Chapters 7–12 point back to that machinery, this pack carries the pointer only.

Not the book's Appendix C runbook. Chapter 10's runbook logic — triage, the gateway containment moves, evidence preservation, eradication, staged recovery — ships in Template 22. The book's full printable Appendix C runbook is a separate artifact, outside this pack's source range, and is not included; the listing promises nothing about it beyond this sentence.

Not a compliance certification, and not legal advice. The crosswalk in Templates 27 and 39 is the book's own mapping, at the granularity the book prints (framework functions and clause families, not clause-by-clause certification mapping). The EU AI Act classification worksheet structures the exercise the book prescribes. Neither certifies compliance with NIST AI RMF, ISO/IEC 42001, OWASP, or the EU AI Act — the book's own position, carried in full: verify obligations against the authoritative texts and qualified legal counsel before any compliance decision.

No runnable code or tooling. The gateway, policy-engine, screening, and detection shapes are specified, audited, and operated through these templates. No working gateway, classifier, policy engine, scanner, or detection pack ships here.

No vendor or product guidance. The vendor assessment (Template 5) turns a provider's marketing into contractual, testable commitments; it selects, installs, and configures no security product.

No guarantee of security. The book's own position: no checklist makes a language-model system "secure". These templates bound, test, and price risk — the red-team gates exist precisely because defenses are measured, never assumed.

Not agent security in depth. Agent identity, permission matrices, sandboxes, and delegation are the companion Securing AI Agents packs' territory, from their own dedicated book. Agents appear here only where Chapters 7–12 treat them: plug-in governance, the agency attack suite, incident containment, and program metrics.

Companion products

AI Security in Production — Part 1: Threat Modeling, Injection, Data & Identity Templates is this pack's sibling, built from Chapters 1–6 of the same book: a companion for when you want the first half of the discipline — never a requirement for this pack to work. (The two parts are priced separately: Part 1 at $39, this Part 2 at $49.) The AI Security Templates & Checklists pack is the cross-book companion — the MCP and enterprise-agent security slice. Securing AI Agents — Parts 1 & 2 are the agent-security companions. This pack stands alone.

Note: the $99 Architecture Templates Bundle currently contains Products #1–#3 only. This pack launches standalone; no bundle inclusion is promised for this product.

Built by the author of AI Security in Production: The Enterprise Defense Handbook, from the defense discipline behind production enterprise AI systems.

About the author

Naveen Kumar Josyula is the author of 35 AI books on Amazon KDP and the CEO and founder of AISoltrics Technologies Inc. He is the author of the source book, AI Security in Production: The Enterprise Defense Handbook, from which every template in this pack is distilled.

Support promise

If anything in this file is wrong or unclear, email support@aisoltrics.com before you judge it — we fix verified issues within one business day and send you the corrected file free. (You can also reach the founder's office at aadhya@aisoltrics.com.) Updates to this product are free to past buyers via your Payhip library. Sold by AISoltrics Technologies Inc.