Written for Security, risk, technology and operations leaders, and the sponsors who have to fund a first security-and-compliance engagement around a system that is already running.
What's in this playbook
This playbook is the operational document behind AsymiLink Meta's Security & Compliance consulting page. It is complementary to LLM Security in Enterprise Environments, already in this library, and is not a replacement for it. That white paper remains the design framework: twelve tests, eight threat classes described qualitatively, defenses as architecture, and a refusal to provide payloads, procedures, or a security certificate. This playbook will not reprint those twelve tests or those eight classes as a method, and it will not give you an exploit, a proof of concept, or a step-by-step attack procedure.
What it is, instead, is the engagement path: six families of security-and-compliance work, fourteen design tests, and a first case consistent with that white paper and with the Enterprise AI Governance Framework. The honest default is to put every running system on an inventory, with an operator who is not only the builder, a stop and a trail. The honest first case is often read-only. A certification that a system is secure, a twelve-week security framework, or an unsupervised high-consequence assistant still off the inventory, is not a first case this playbook will recommend. A person sits on every write and every customer-facing message. Hosting is decided before the case is funded. Where data cannot leave, AsymiLink AI runs on your own hardware, with an optional air gap. Isolation is not coverage of a design that was never finished, and it is not a close of any named regime.
How the tests and the method are meant to be used
The financial method is The Complete Guide to AI Automation ROI. This playbook will not invent a security-sector return or a "cost of a breach" figure, and it will not promise a twelve-week go-live. It will not close a named regime, will not offer a "flat-fee security" product as a new commitment, and will not certify that a system is secure. 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 security project, and is not used as a sector benchmark. Built to SOC 2 standards is the exact wording we use; it is not a certification of the reader's system, and not a finding that a system is safe to run unattended.
The illustrative firm is the same hypothetical community bank as in the LLM Security white paper, the Enterprise AI Governance Framework, the financial-services paper, and the RAG guide: overall 22, Developing, with a next-step target of 25, still Developing. Its model-risk practice is a strength; two generative-AI assistants are off the inventory, and have no operator who is not the builder. The honest first work is to put those assistants on the inventory, name the threat classes qualitatively, and give them an operator, a stop and a trail. A companion workbook, asymilink-security-compliance-playbook.xlsx, holds a family register, the fourteen tests, questions for counsel, a baseline sheet, and the same illustrative bank. Do not wait to be Leading.
Six families of security-and-compliance work
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 and an operator on every running system (first)
F2
A bounded, often read-only, first case, with threat classes named qualitatively
F3
Access, identity and least privilege
F4
A trail that can reconstruct a day (later)
F5
A customer-facing or write-back security design (later)
F6
Hosting, and what would leave, decided before the case is funded
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 design tests (excerpt)
Fourteen design 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
including a named inventory that covers unsanctioned use, a person on every write, counsel asked in time without a close of a named regime, and a refusal to provide a payload or to certify that a system is secure.
A completed test sheet is not a certificate, not a compliance score, and not a finding that a system is safe or aligned with 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.