Written for Control owners, security and technology leaders, and the sponsors who have to assemble evidence for AI systems that are already running.
What's in this checklist
This checklist is the operational document for SOC 2 evidence on AI systems. It is complementary to the LLM Security white paper and the Enterprise AI Governance Framework already in this library, and is not a replacement for them. Those documents remain the companions for security tests and for governance. This checklist will not reprint their tests, and will not provide payloads. It is evidence only. It is not a certification of your system, and not a promise that we will get the reader certified.
What it is, instead, is an evidence checklist: six families of evidence work, fourteen evidence tests, and a first-case path that assembles evidence for the AI systems already running — inventory, access, change control, logs, vendors, and hosting — in a form a reviewer could use. A completed sheet is not a finding of certification or readiness, and is not a substitute for a named regime. Built to SOC 2 standards is how we build, not a certification of your system. We will not get the reader certified. We will not promise a certified result. A twelve-week "get certified" program is not a first case this checklist will recommend. This checklist will not invent a pass rate or a "readiness score," and will not close a named regime.
How the evidence tests and the method are meant to be used
The financial method is The Complete Guide to AI Automation ROI, where a process is being funded. This checklist will not invent a SOC 2-sector return, and it will not promise a twelve-week certification. The 3.8x average first-year ROI across client engagements is an average, not a forecast or a commitment for any individual project. ROI is measured across client engagements against a baseline agreed before work starts. It is an average, not a forecast for any SOC 2 project, and is not used as a sector benchmark.
The illustrative firm is a new hypothetical mid-size company with AI systems already running, introduced in this checklist. It is not the community bank already used in the LLM Security white paper, and not the mid-size insurer already used in the governance framework. Overall 18, Emerging, Data Foundation the constraint at 2, Infrastructure and Governance relative strengths at 3, with a next-step target of 24, Developing. Those strengths are not a finding of readiness. Evidence for access, change control, logs, and vendors has not been assembled in a form a reviewer could use. The honest first work is to assemble that evidence for the systems already running, not to fund a twelve-week "get certified" program. The target is a next level, not a timeline, not a promise, and not a certification. A companion workbook, asymilink-soc2-readiness-checklist.xlsx, holds a family register, the fourteen evidence tests, questions for counsel, a baseline sheet, and the same illustrative company. Do not wait to be Leading. Do not treat a completed sheet as a finding of readiness.
Six evidence families
each with an honest first case, a case that is too wide, and a gate. Each item below is the shape of the map; the full wording, honest first cases and “too wide” warnings are in the PDF.
F1
Inventory evidence for every running AI system (first)
F2
Access and identity evidence
F3
Change-control evidence
F4
Logs and a trail that can reconstruct a day
F5
Vendor and subprocessor evidence
F6
Hosting evidence, and what would leave. Isolation is not a certification.
The PDF gives, for each family, an honest first case, a case that is too wide, and a defined gate. The summaries above are the shape of the map, not a substitute for that wording.
Fourteen evidence tests (excerpt)
Fourteen evidence tests are applied to every case. The answers will differ. The standard does not. The full wording is in the PDF and the companion workbook.
T1–early
a named inventory in a form a reviewer could use; a trail that can reconstruct a day; counsel asked without a date taken from this document.
Mid-tests
a refusal to treat a completed sheet as a finding of readiness or to get the reader certified; evidence for access, change control, logs, vendors and hosting.
Later tests
a signed baseline on the ROI method where a process is being funded; isolation is not a certification.
A completed sheet is not a finding of certification or readiness, and is not a substitute for a named regime.
Get the guide and workbook
Enter your name, work email and company. The files appear on this page immediately. A backup download link is also emailed when email delivery is available.