START HERE
What you will learn
- Distinguish a design principle from a risk list or verification standard.
- Choose a reference that matches the decision you need to make.
- Connect an intended safeguard to evidence rather than a compliance slogan.
Before you start: No security background required. Think about one application, integration or technical decision you own.
Reference edition: OWASP Top 10:2025, ASVS 5.0.0 and NIST CSF 2.0. Sources checked 2026-11-11.
Start with the question you need answered
“Should this automation have administrator access?” is a design question. Least privilege helps answer it: give an identity only the permissions needed for its task. “Which application failures should we examine?” is a different question, for which a risk-awareness list can be useful.
OWASP is a foundation that maintains several projects, not one interchangeable checklist. Its web-application Top 10 and its risk list for large language model (LLM) applications have different scopes. The Application Security Verification Standard, or ASVS, provides more detailed security requirements. The US National Institute of Standards and Technology (NIST) publishes the Cybersecurity Framework, or CSF, which organises cybersecurity outcomes across responsibility, protection and incident handling.
A reference is not evidence that your system is secure
Naming a framework does not show that a control exists or works. For a document portal, “access control is important” is a principle-level statement. “A user cannot retrieve another customer’s document, including through its software interface” is a testable expectation. A recorded test using isolated test accounts is evidence about that expectation.
Keep the application version, scope, selected requirements and unresolved gaps with the evidence. These guides help explain and apply references; they do not constitute a certification, legal determination or assessment of your actual system.
Choose the reference by the job it does
| Kind of reference | Question it helps answer | Example | Useful output |
|---|---|---|---|
| Design principle | How should this feature be constrained? | Least privilege | A permission boundary |
| Risk-awareness list | What important failure types might we overlook? | OWASP web or LLM risk lists | A set of investigation questions |
| Verification standard | What technical behaviour should we check? | OWASP ASVS | Scoped requirements and test evidence |
| Risk-management framework | Who decides, prioritises and follows up? | NIST Cybersecurity Framework | Owned outcomes and improvement decisions |
Step-by-step process
Describe the system and decision
State what you are building, whose information it handles and the action you need to authorise. Keep the first scope narrow enough to inspect.
Choose relevant principles
For an integration, start with required permissions, what should happen when a check fails and how a consequential action receives approval.
Use a risk list to ask better questions
Look for risks that apply to the actual design. An AI assistant that can call tools needs questions about data access and permitted actions as well as ordinary web security.
Turn concerns into checks
Select appropriate verification requirements and write the expected result. Identify who can test the behaviour and where the evidence will be recorded.
Manage gaps and revisit changes
Assign an owner and next decision to unresolved risks. Revisit the scope when you add data, users or a new tool permission. A previous review does not automatically cover the changed system.
FOLLOW ONE EXAMPLE
Worked scenario
A stakeholder reviewing a customer-document assistant
In this fictional project, an assistant searches customer files and drafts summaries. The stakeholder needs to decide what evidence is necessary before a limited trial.
- Least privilege leads to a read-only search identity restricted to the requesting user’s permitted documents.
- An OWASP-informed review asks about unauthorised records, untrusted document instructions and unsafe handling of generated output.
- The implementation review selects concrete access checks, while the project owner records unresolved risks and decides whether the trial may proceed.
What this shows: The references contribute different parts of one decision. The evidence concerns defined behaviour in the proposed trial, not an unsupported claim that the assistant is “OWASP certified”.
The process at a glance
Read Mermaid diagram source
flowchart LR A[Describe the system] --> B[Choose design principles] B --> C[Identify relevant risk questions] C --> D[Define checks and evidence] D --> E[Owner decides on remaining risk] E --> F[Revisit when the system changes]
What you should have at the end
- A clear reason for choosing each reference.
- At least one security expectation that someone can verify.
- A named owner and decision for an unresolved gap.
Common mistakes and how to avoid them
Treating every reference as a checklist
A governance framework and an application verification standard do different jobs.
Using a project name as a security badge
Describe the scope, checks and remaining limitations instead.
Starting with every possible standard
Begin with the system and its consequences, then choose the relevant depth.
APPLY THE IDEA
Try it yourself
Your task: Choose one proposed integration. Write one design principle, one risk question and one technical check relevant to it.
What to produce: A short decision note connecting the concern, control, evidence and accountable owner.
Check your understanding: Could another stakeholder tell the difference between your intention and evidence that it works?
Key terms in plain English
- Principle
- A general rule that guides choices, such as limiting permissions.
- Control
- A measure intended to prevent, detect or reduce a risk.
- Verification
- Checking a defined behaviour and recording the result.
- Residual risk
- Risk that remains after the selected measures are considered.
- Risk
- The possibility of an unwanted outcome, considering its likelihood and consequences.