Skip to main content
Lightbridge.ai

Back to the CIO and CSO starting point Security and GRC hub

RL Written by Robert LabardeeFounder and CEO

AI security questionnaire resources: public standards and templates

AI security questionnaire resources are the public standards, control catalogs, and vendor-assessment templates that questionnaires draw on. Lightbridge indexes fourteen entries here, from the CSA AI Controls Matrix to NIST AI 600-1 and the OWASP LLM Top 10, with each publisher's version, scope, stated terms, and official link.

This page links out to the official sources and does not reproduce their text. Landing pages and PDFs were read on 2026-11-03; the AICM downloads and the machine-readable bundle were not opened. The table states what each publisher says about its own terms. Lightbridge does not endorse a vendor here. For the process, see Lightbridge's guide to answering AI security questionnaires. For a starting point, see the Lightbridge questionnaire templates.

The public sources Lightbridge indexes, with publisher, version, scope, and terms.

Fourteen entries (thirteen distinct pages, since the AI-CAIQ ships inside the AICM download), grouped loosely from control catalogs to legal texts. The terms column repeats what the publisher states. Where a page states nothing, the cell says so. A missing licence statement does not mean permission.

Resource and publisher Version or date What it is and covers Terms as the publisher states them Official link
CSA AI Controls Matrix (AICM) Cloud Security Alliance v1.1, released 2026-06-22. CSA states v1.1 is the newest version; v1 (243 objectives) is the earlier release. A control catalog. CSA states v1.1 has 247 control objectives across 18 security domains. CSA also publishes role-specific guidelines for five roles: Model Provider, Orchestrated Service Provider, Application Provider, AI Customer, and Cloud Service Provider. The page says the matrix is freely available to download and shows login links. It states no licence or reuse terms, so reuse terms are unconfirmed. CSA AICM v1.1
CSA AI-CAIQ Cloud Security Alliance Included in the AICM v1.1 download (2026-06-22). CSA also lists a machine-readable bundle (JSON, YAML, OSCAL) of controls, AI-CAIQ, and mappings, released 2026-08-04. A question set that maps to the AICM, which CSA describes as intended for self-assessment or third-party vendor evaluation. The download also includes guidance on filling it in and a STAR for AI Level 1 submission guide. Same page, same position: no licence or reuse terms stated. Not stated in the documentation we reviewed. CSA AICM v1.1 download page
NIST AI 600-1, Generative AI Profile NIST Published 2024-07-26. DOI 10.6028/NIST.AI.600-1. A profile of the AI RMF for generative AI. Section 2 names 12 risks, and suggested actions carry IDs such as GV-6.1-007 (third-party inventory) and GV-6.1-009 (procurement due diligence). The PDF says it is available free of charge and points to the NIST Technical Series Policies for copyright, use, and licensing. We could not retrieve that licence text. NIST AI 600-1 PDF
NIST AI Risk Management Framework 1.0 NIST Released 2023-01-26 for voluntary use. NIST states it is being revised as part of the White House AI Action Plan (page modified 2026-08-13). Four core functions: Govern, Map, Measure, and Manage. Its subcategories include Govern 6.1, which covers third-party policies. Not stated in the documentation we reviewed. NIST AI RMF page
OWASP Top 10 for LLM Applications 2025 OWASP Gen AI Security Project 2025 list (LLM01:2025 to LLM10:2025). The cited page was modified 2025-04-17. Ten risk categories, from LLM01 Prompt Injection and LLM02 Sensitive Information Disclosure to LLM06 Excessive Agency and LLM10 Unbounded Consumption. The site footer states a Creative Commons Attribution-ShareAlike 4.0 licence for site content unless otherwise specified. OWASP LLM01 page
MITRE ATLAS MITRE The atlas-data repository shows release v2026.10. A public knowledge base of adversary tactics and techniques against AI systems. The atlas.mitre.org home page shows 16 tactics, 221 techniques, 44 mitigations, and 77 case studies. Technique IDs use the AML.Txxxx form. Not stated in the documentation we reviewed. The repository shows a LICENSE file, which we did not read. MITRE ATLAS data on GitHub
HECVAT 4 EDUCAUSE EDUCAUSE lists version 4.1.6 as current. The page is undated. A higher-education vendor assessment toolkit. EDUCAUSE says HECVAT 4 added privacy and AI questions. The question IDs sit in the downloadable spreadsheet. EDUCAUSE owns the copyright. Colleges and vendors may use it for college and university business at no cost. All other parties, including third-party risk management platforms, need specific permission and a licence. Trademark terms: not stated in the documentation we reviewed. EDUCAUSE HECVAT page
GovAI Coalition AI FactSheet GovAI Coalition, hosted by the City of San Jose Page undated. An AI FactSheet (PDF and DOCX), a FactSheet for service providers, a Vendor Agreement, and a Buyer's Guide to Enterprise GenAI Tools. The coalition states the templates align with the NIST AI RMF. The page invites agencies to borrow, edit, and adopt the templates and carries a not-legal-advice note. It states no formal licence, and commercial reuse is unconfirmed. GovAI Coalition templates
WaTech generative AI procurement guidelines Washington State Office of the CIO (WaTech) September 2024, issued under Executive Order 24-01. Appendix A is a sample generative AI procurement questionnaire and Appendix B is a sample initial use case assessment. No licence sentence found in the fetched PDF. WaTech guidelines PDF
K-12 Gen AI Readiness Checklist Council of the Great City Schools and CoSN Version 1.3, February 2026. Seven sections, from Executive Leadership Readiness (24 items) to Security Readiness (7) and Legal/Risk Management (12). Procurement sits under Operational Readiness. The PDF says a Creative Commons licence is not yet finalized. CoSN checklist PDF
EU AI Act, Regulation (EU) 2024/1689 European Union, via EUR-Lex In force 2024-08-01. The Act has since been amended, so check the consolidated text on EUR-Lex before citing any article or date. Regulation (EU) 2026/1744 (8 July 2026) amended the Act, so Art. 4 on AI literacy now reads as measures to support the development of AI literacy, with no guaranteed level. Read the consolidated text (2026-07-27). Other articles relevant to vendor questions include Art. 25 (responsibilities along the value chain), Art. 26 (deployer obligations), and Art. 50 (transparency). No reuse licence found. Reuse terms unconfirmed. EUR-Lex, Regulation 2024/1689
GDPR Article 28 (processor) European Union, via EUR-Lex Regulation (EU) 2016/679, applicable from 2018-05-25. The controller uses only processors with sufficient guarantees, and sub-processors need written authorisation. The processor contract covers items (a) to (h), from documented instructions to information and audits. No reuse licence found. Reuse terms unconfirmed. EUR-Lex, Regulation 2016/679
ISO/IEC 42001:2023 ISO and IEC Edition 1, published 2023-12. Requirements for an AI management system. The ISO catalog page lists 51 pages and a price of CHF 225, and does not list the Annex A control themes. For explainers, see Lightbridge Automation's ISO/IEC 42001 guide on lightbridgeautomation.com. Paid standard. The catalog page does not state reuse terms. Not stated in the documentation we reviewed. ISO catalog page
SIG (Standardized Information Gathering) Shared Assessments Page modified 2026-06-04. Shared Assessments states SIG EV covers 21 risk domains, including Artificial Intelligence, and maps to ISO/IEC 42001:2023 and NIST AI 100-1 (for explainers on the standard, see lightbridgeautomation.com). Question text is not public, only the domain list. A paid product. Shared Assessments lists a corporate licence at $7,000 per year, also included with membership. Shared Assessments SIG page

Landing pages and PDFs read 2026-11-03; AICM downloads and the machine-readable bundle not opened. Lightbridge is not a law firm and this page is not legal advice. Confirm versions and terms on the publisher's own page before you rely on them.

How Lightbridge suggests using these sources, and what to leave alone.

Cite identifiers, such as GV-6.1-007 or LLM06, and link to the publisher. Check terms before copying anything: the HECVAT is the clearest case, because EDUCAUSE states that parties outside colleges and vendors need a licence, including third-party risk platforms. Then answer from your own evidence. A standard tells you what to ask. It cannot tell you what is true about your systems.

Some sources are thin. The AICM landing page does not list domain IDs, and we did not open the download or the machine-readable bundle. SIG question text is not public, and several other sources state no licence. Lightbridge shows those gaps instead of filling them. The sibling pages cover the rest of the picture: AI contracts and legal terms, observability and audit logging, and AI access paths. This page is part of Lightbridge's security and GRC advisory hub.

How Lightbridge advises on AI security questionnaires.

Lightbridge is an independent advisory and program-management firm. We help a CIO or CSO build an evidence library, pick which public sources a questionnaire program should align to, and run the program through delivery. Where hands-on work is needed, vetted partners deliver under Lightbridge's program management.

Lightbridge Automation, a sibling practice in the Lightbridge group, publishes explainers on the NIST AI RMF, the EU AI Act, and zero data retention, and offers consulting on governance. If a recommendation here points to work a sibling practice would deliver, Lightbridge discloses that upfront and you remain free to take the work elsewhere.

AI security questionnaire resources: questions Lightbridge hears

Which public resource should we read first?

Start with what your customers cite. If buyers send a CSA AI-CAIQ or a HECVAT, read that first. If they send a custom list, NIST AI 600-1 has action IDs GV-6.1-007 and GV-6.1-009, which cover third-party inventory and procurement due diligence.

Can we copy questions from these standards into our own template?

Check each publisher's terms first. EDUCAUSE states that parties other than colleges and vendors need specific permission and a licence to use the HECVAT. OWASP states a Creative Commons Attribution-ShareAlike licence for site content. For CSA, GovAI Coalition, WaTech, and MITRE, the pages we reviewed state no clear reuse terms. Lightbridge cites identifiers and links out. It does not reproduce publisher text.

Is the CSA AI-CAIQ free?

CSA states the AICM download, which includes the AI-CAIQ, is freely available, and the page shows login links. It states no licence or reuse terms. Lightbridge reads that as free to download and unconfirmed for reuse.

Does the SIG publish its AI questions?

No. Shared Assessments states SIG EV is a paid product, and the question text is not public. The public page lists the domains, including Artificial Intelligence. A template can cite the domain name only.

Does listing the AI management system standard mean Lightbridge holds it?

No. The entry is a reference to a public standard that questionnaires may mention. It says nothing about Lightbridge itself.

Where do EU AI Act and GDPR fit in a security questionnaire?

They arrive as legal and contract questions. GDPR Article 28 sets what a processor contract must cover, and the EU AI Act sets duties for providers and deployers. The EU AI Act has been amended, so confirm article text and dates against the consolidated version on EUR-Lex. Lightbridge is not a law firm. Have counsel confirm the position.

How current is this list?

Landing pages and PDFs were read on 2026-11-03; the AICM downloads and the machine-readable bundle were not opened. Each row shows the version or date the publisher states. Some pages are undated, and the table says so. Standards change, so check the publisher's page before you cite a version.

Bring the questionnaire: talk to a Lightbridge advisor.

Send the questionnaire your buyers use. Lightbridge will help you map it to public sources and the evidence you already hold.