Web application security policy template for a small team
A modular policy for one web application, written for a business with a small team and an outside developer.
Every clause names an owner, the evidence that shows it is followed and when it is reviewed, and a filled-in exception register shows how gaps are tracked. It is a starting point for your own decisions, not a certificate.
On this page
Modular web application security policy with exception register
Copy the text into your own document, delete the modules and clauses your team will not follow, and replace every prompt in round brackets with your own names, numbers and record locations. Each clause carries owner, evidence and review fields, and some carry an exceptions line.
How to use itHide the instructions
The example register at the end uses an invented business to show the format; replace it with your own entries before the policy is approved.
security-policy-template.txt
WEB APPLICATION SECURITY POLICY
Application: (name of the application and its main web address)
Business: (legal name of the business)
Policy owner: (name, role)
Approved by: (name, role) on (date)
Version: (number)
Next review due: (date)
HOW TO USE THIS POLICY
This policy covers one web application and the people and suppliers who can change it. It is an internal starting point for the business's own decisions. It is not a certificate, and adopting it does not show that the application or the business meets any framework, standard or law.
- Delete every module and clause the business will not follow. A clause nobody follows does more harm than a missing one, because readers take the policy as a description of what the business does.
- A clause the business intends to follow but cannot meet today stays in, with an entry in the exception register (section 9).
- Replace each prompt in round brackets with a name, a number or the location of a record.
- Every clause ends with Owner, Evidence and Review, and some carry an Exceptions line too. Owner: the named person accountable for it. Evidence: the record that shows it is followed, and where that record is kept. Exceptions: the route for departing from it; a departure from a clause without this line goes in the exception register (section 9). Review: when it is next checked.
- Lines starting "See:" point to source wording the clause relates to, so a developer or reviewer can read it. They do not state that the clause, or the application, meets that source. ASVS references use the format v5.0.0-(chapter).(section).(requirement).
1. SCOPE AND ROLES
1.1 In scope: (application name), its production and (staging or test) environments, its code repository at (location), its hosting with (provider), and these third-party services that hold its data or run its code: (list).
1.2 Out of scope: the office network, staff devices and staff email, which are covered by (document name, or "not yet covered").
1.3 Roles
Policy owner: (name)
Technical lead: (name, and whether in-house or a supplier)
Incident contact: (name, phone)
Incident deputy: (name, phone)
Exception approver: (name, role)
2. MODULE A: ACCESS
A1 Every account with administrator rights to the application, its hosting, the domain registrar, DNS, the code repository, the database and the payment provider belongs to one named person. Administrator logins are not shared.
Owner: (name)
Evidence: the account list exported from each system on (date), stored at (location)
Exceptions: a shared login that cannot be split yet goes to the register
Review: (how often), next on (date)
A2 Multi-factor sign-in is turned on for every account in A1 wherever the system offers it.
Owner: (name)
Evidence: the multi-factor status shown in each system's user list, captured at each A4 review
Exceptions: a system that does not offer it goes to the register, with what is used instead
Review: with A4
A3 The roles inside the application are written down: what each role can see and change, and how one customer's records are kept apart from another customer's.
See: OWASP ASVS v5.0.0-8.1.1
Owner: (technical lead)
Evidence: the role table at (location), updated in the same change as any change to roles or permissions
Exceptions: register
Review: whenever a role is added or changed
A4 Access is granted at the lowest level a person needs to do their work, approved by (role), and reviewed (how often). At each review every account is marked kept, reduced or removed, with the reviewer's name and the date.
See: NIST CSF 2.0 PR.AA-05
Owner: (name)
Evidence: the dated review record at (location)
Exceptions: register
Review: (how often), next on (date)
A5 When someone leaves or changes role, including supplier staff, their access to every system in A1 is removed or reduced within (number) business days, and any shared password they knew is changed. Disabling their application account also ends its active sessions.
See: OWASP ASVS v5.0.0-7.4.2 (ending sessions when an account is disabled)
Owner: (name)
Evidence: the leaver checklist, with the date access was removed from each system
Exceptions: register
Review: after each leaver
3. MODULE C: CHANGE MANAGEMENT
C1 Changes to code and configuration reach production only through (code repository) and (deployment method). Files are not edited directly on the production server.
Owner: (technical lead)
Evidence: the deployment history in (system)
Exceptions: an emergency change made outside this route is written up in (location) within (number) days, with the reason
Review: (how often)
C2 A second person reviews each change before release. Where one developer works alone on the application, the change record says what changed and why, and (name) reads the release notes before the release goes out.
Owner: (technical lead)
Evidence: approvals or comments on merge requests in (repository), or the release notes signed off by (name)
Exceptions: register
Review: (how often)
C3 Changes that touch sign-in, roles or permissions, payments, personal information, file uploads or a new integration are labelled security-relevant in (tracker), and each labelled change is considered under clause T1.
Owner: (technical lead)
Evidence: the labelled changes in (tracker)
Review: (how often)
C4 Backend secrets, such as API keys and database passwords, are kept in (secrets store or hosting environment settings). They are not stored in the code repository, in build output or in chat messages.
See: OWASP ASVS v5.0.0-13.3.1 (secrets store, source code and build output)
Owner: (technical lead)
Evidence: a list of each secret and where it is held, at (location)
Exceptions: register
Review: (how often), and straight after any suspected exposure
C5 Debug modes are switched off in production, and test code, sample content and development-only features are not deployed to it.
See: OWASP ASVS v5.0.0-13.4.2 and v5.0.0-15.2.3
Owner: (technical lead)
Evidence: the release checklist item, ticked for each release
Review: each release
4. MODULE D: DEPENDENCIES
D1 An inventory of third-party components is kept: code libraries (the package manifest and lock files in the repository, which usually list nested dependencies too), plugins and themes, and scripts the browser loads from other domains, such as analytics, chat and payment widgets.
See: OWASP Top 10:2025 A03 Software Supply Chain Failures
Owner: (technical lead)
Evidence: the lock files in (repository), and the plugin and script list at (location)
Review: whenever a component is added or removed
D2 A component with a published vulnerability is updated or removed within these time frames, set by severity: critical (number) days, high (number) days, all others (number) days. Other updates are applied at least every (period).
See: OWASP ASVS v5.0.0-15.1.1 and v5.0.0-15.2.1
Owner: (technical lead)
Evidence: the advisory alerts and the date each was closed, in (system)
Exceptions: a component past its time frame goes to the register
Review: (how often)
D3 Advisory alerts for the components in D1 go to (named person or shared inbox). Each alert is triaged within (number) business days: update, remove, or write down why it does not apply.
Owner: (name)
Evidence: the alert history in (system), with the triage decision for each alert
Review: (how often)
D4 New components come only from (official package registry, plugin directory or vendor site). A component that is no longer maintained is replaced, or recorded in the register with a date.
See: OWASP Top 10:2025 A03 Software Supply Chain Failures
Owner: (technical lead)
Evidence: the change record for each new component
Review: whenever a component is added
5. MODULE L: LOGGING
L1 A log inventory lists each log the application and its services produce (for example the application log, web server log, hosting control panel login history and payment provider activity), what it records, where it is kept, who can read it and how long it is kept.
See: OWASP ASVS v5.0.0-16.1.1
Owner: (technical lead)
Evidence: the inventory at (location)
Review: whenever a service is added or removed
L2 The application logs successful and failed sign-in attempts and refused access attempts, with the time, the account and where the request came from.
See: OWASP ASVS v5.0.0-16.2.1, v5.0.0-16.3.1 and v5.0.0-16.3.2
Owner: (technical lead)
Evidence: a sample log entry for each event type, captured at each review
Exceptions: register
Review: (how often)
L3 Passwords and payment card details are not written to logs, and session tokens are left out or masked.
See: OWASP ASVS v5.0.0-16.2.5
Owner: (technical lead)
Evidence: sample log entries checked at each review
Review: (how often)
L4 Logs are kept for (period) in (location), where the application's own administrator accounts cannot change or delete them.
See: OWASP ASVS v5.0.0-16.4.2
Owner: (name)
Evidence: the retention setting in the hosting or logging service, captured at each review
Exceptions: if the hosting plan cannot do this, the gap goes to the register
Review: (how often)
L5 These alerts reach a person: (list, for example repeated failed administrator sign-ins, a new administrator account, a payment provider fraud alert, a hosting suspension notice). Each goes to (name or inbox), and whoever receives it decides within (time) whether to escalate under module I.
Owner: (name)
Evidence: the alert settings, captured at each review, and the most recent alert received with what was done about it
Review: (how often)
6. MODULE I: INCIDENT ESCALATION
I1 Anyone who suspects a security problem, whether a staff member, a supplier or a customer, tells the incident contact by phone on (number). If there is no answer within (minutes), they call the incident deputy.
See: NIST CSF 2.0 RS.MA-04
Owner: (incident contact)
Evidence: this clause in (staff handbook or onboarding notes), and the contact sheet in I4
Review: (how often), and after each incident
I2 Until the hosting provider or a specialist advises otherwise, nobody deletes files, accounts or records, restores a backup over the live site, or cancels the hosting plan. The first steps are to start a written log with times and names, call the hosting provider on the number in the hosting account, and ask for the files and logs to be kept as they are.
See: NIST CSF 2.0 RS.AN-06 (recording the actions taken)
Owner: (incident contact)
Evidence: a blank incident log kept at (location)
Review: with I1
I3 The response itself is carried out by (incident response provider or website cleanup specialist, with contact details), engaged by (role). If the business holds cyber insurance, what the insurance policy says about notifying the insurer and choosing a responder is checked first: (insurer, policy number, claims line).
Owner: (name)
Evidence: the agreed provider and insurer details on the contact sheet, checked on (date)
Review: (how often)
I4 An offline contact sheet lists: hosting provider support and account ID, domain registrar, developer or agency, incident response provider, insurer, payment provider, and a lawyer who handles privacy matters. A printed copy is kept at (location).
Owner: (incident contact)
Evidence: the sheet, dated when last checked
Review: (how often)
I5 Decisions about telling customers, the insurer, police or a regulator are made by (role) with advice from (lawyer or adviser). This policy names who decides. It does not set out what any law requires.
Owner: (role)
Evidence: each decision recorded in the incident log
Review: after each incident
I6 Within (number) days of the end of a response, (name) writes down what happened, what changed and which clauses of this policy need updating.
Owner: (policy owner)
Evidence: the incident summary at (location)
Review: after each incident
7. MODULE S: SUPPLIER ACCESS
S1 A supplier register lists each supplier that can reach the application's code, hosting, data or admin areas: what it can reach, through which accounts, under which agreement, and who to contact there about security.
See: NIST CSF 2.0 GV.SC-07
Owner: (name)
Evidence: the register at (location)
Review: (how often), and whenever a supplier starts or finishes
S2 Supplier staff use their own named accounts with multi-factor sign-in, under A1 and A2. Nobody gives a supplier the login of an owner or staff member.
Owner: (name)
Evidence: the account lists from A4 reviews, showing supplier accounts by name
Review: with A4
S3 The business, not a supplier, is the account owner of the domain registration, DNS, hosting account, code repository and payment provider. Suppliers are added as users.
Owner: (policy owner)
Evidence: the owner details shown in each account, captured on (date)
Exceptions: register, with a date for the transfer
Review: (how often)
S4 Agreements with suppliers who can change code or reach data state who at the supplier works on the application, how quickly the supplier tells (incident contact) about a suspected incident affecting it, and that access and copies of data are handed back or deleted when the work ends.
See: NIST CSF 2.0 GV.SC-05 and GV.SC-10
Owner: (name)
Evidence: the agreement reference for each supplier in the register
Exceptions: a supplier without a written agreement goes to the register
Review: whenever an agreement is signed or renewed
S5 When a supplier's work ends, its accounts are removed, passwords and keys it held are changed, and its deploy keys and integrations are revoked, on the last day of the engagement.
See: NIST CSF 2.0 GV.SC-10
Owner: (name)
Evidence: the offboarding checklist, with dates
Review: at the end of each engagement
8. MODULE T: SECURITY TESTING
T1 A security test or review of the application is considered, and the decision written down, when any of these happens:
- a new feature handles payments, personal information or file uploads
- sign-in, roles, permissions or the separation of one customer's data from another's changes
- a new integration or a public API is added
- the host, developer or agency changes
- a customer, contract or auditor asks for a recent test
- (technical lead) judges a change labelled under C3 to be significant
Testing frequency, if the business sets one: (frequency, or "triggered by the events above only")
Frequency set by a contract or auditor, if any: (frequency, and the document it comes from)
Owner: (policy owner)
Evidence: the decision record for each trigger, at (location)
Review: (how often)
T2 Each test is authorised in writing by the owner of every target, and the hosting provider's conditions for testing are checked before it starts. Destructive actions are off by default for real people's accounts and records that existed before the test, and reset, lockout and deletion checks run only on test accounts and test records where the rules of engagement allow them.
Owner: (policy owner)
Evidence: the signed authorisation and scope for each test
Review: for each test
T3 Each test has a written scope, and its report states what was tested, the dates, the build or version tested, and what was out of scope or not performed.
Owner: (technical lead)
Evidence: the report, stored at (location)
Review: for each test
T4 Each finding goes into (tracker) with an owner and a target date set by its severity: the D2 time frames for a vulnerable component, and (time frames) for anything else. Fixes are retested, and each finding is recorded as fixed, partially fixed, still reproducible or not retested. A finding marked not retested is not described as fixed.
See: NIST CSF 2.0 ID.IM-02
Owner: (technical lead)
Evidence: the findings list with statuses in (tracker)
Review: until every finding is closed or in the register
T5 If the tester also built or hosts the application, the report and any summary shared outside the business say so, and the work is described as a review by the builder rather than an independent test.
Owner: (policy owner)
Evidence: the report's statement of who built and who hosts the application
Review: for each test
T6 A test report leaves the business only with approval from (role), and a full report only under a confidentiality agreement.
Owner: (policy owner)
Evidence: a record of who received each report, and when
Review: (how often)
T7 Security reports from outside the business arrive at (published contact, such as the security address in a vulnerability disclosure policy) and are triaged by (name) within (number) business days.
See: NIST CSF 2.0 ID.RA-08
Owner: (name)
Evidence: the published contact, and the triage record for each report
Review: (how often)
9. EXCEPTION REGISTER
Each exception records:
ID
Clause
What is not met, and where
Why
Accepted by (name, role; where the team allows it, not the person who raised it)
What reduces the risk in the meantime ("nothing" is allowed, but it is written down)
Raised (date)
Expires (date)
Status (open; renewed, with the new reason and date; or closed, with the date and the evidence that closed it)
Every exception has an expiry date. On that date it is closed, or renewed with a new reason and a new date. Closed exceptions stay in the register.
EXAMPLE REGISTER (ILLUSTRATION ONLY)
The business, people, systems and dates below are invented to show the format. Replace them with your own entries before the policy is approved.
Example business: an online booking service with about 25 staff, a customer portal built and maintained by an outside agency, and card payments taken on the payment provider's hosted page.
Register as at: 1 October 2026
EX-01
Clause: A1 (named administrator accounts)
Not met: the payment provider account has one administrator login, shared by the office manager and the bookkeeper, because the current plan allows one user.
Why: the plan upgrade was due at the next billing review.
Accepted by: Managing director
Meanwhile: multi-factor sign-in is on, with codes going to a phone kept in the office, and the password is changed whenever someone who knows it leaves.
Raised: 3 February 2026
Expires: 30 June 2026
Status: Closed 12 June 2026. Plan upgraded, two named accounts created and the shared login disabled. Evidence: user list export dated 12 June 2026.
EX-02
Clause: D2 (critical update within the set time frame)
Not met: a date-picker library in the customer portal has a published critical advisory, and the fixed version needs a code change the agency could not ship within the time frame.
Why: the fix changes how bookings are displayed and has to be tested with the calendar integration.
Accepted by: Managing director
Meanwhile: the affected feature, calendar export, is switched off for customers until the update ships.
Raised: 14 April 2026
Expires: 15 May 2026
Status: Closed 9 May 2026. Update released, calendar export switched back on, and the agency's retest recorded the finding as fixed. Evidence: change record and retest note.
EX-03
Clause: L4 (logs kept where administrators cannot change them)
Not met: the hosting plan keeps web server logs for 14 days and cannot send them anywhere else.
Why: a plan with log forwarding is part of the hosting renewal.
Accepted by: Operations manager
Meanwhile: nothing beyond the 14 days the host keeps. Recorded so the gap is known if an incident is noticed late.
Raised: 2 March 2026
Expires: 31 March 2026
Status: Renewed 31 March 2026 with a new expiry of 30 November 2026, because the hosting renewal moved to November. A second renewal needs the managing director's approval.
EX-04
Clause: T1 (test considered before a feature that handles file uploads)
Not met: customers can now upload documents to their bookings. The feature launched before the test decision was recorded, and a test is booked for 20 October 2026.
Why: the launch date was tied to a partner promotion.
Accepted by: Managing director
Meanwhile: uploads are limited to PDF and image files, and each upload is shown only to the customer who uploaded it and to staff.
Raised: 18 August 2026
Expires: 31 October 2026
Status: Open.
EX-05
Clause: S3 (the business owns the domain and DNS accounts)
Not met: the domain is registered to the agency that built the first version of the site.
Why: the transfer needs the agency's authorisation code, requested on 10 September 2026.
Accepted by: Operations manager
Meanwhile: the registrar account has multi-factor sign-in, and the agency has confirmed in writing which of its staff can access it.
Raised: 10 September 2026
Expires: 31 October 2026
Status: Open.What a web application security policy is for
A web application security policy is the written record of how a business looks after one application: who can reach its admin areas and data, how changes get to production, how third-party code is kept current, what is logged, who is called when something looks wrong, what suppliers may touch, and what prompts a test.
It is narrower than an information security policy for the whole business. It covers the customer portal, booking system or online store itself, and the people and suppliers who can change it.
The template is written for a business with $1M to $50M in annual revenue: a small team, one main web application and, usually, an outside developer or agency. That shape decides what the policy can ask for. There is often no dedicated security team, so every clause has to fit into work someone already does.
A policy has three kinds of reader. Your team reads it to know who decides what. Your developer or agency reads it to know what you expect of them. A customer's reviewer or an auditor reads it as a list of statements and then asks for the records behind them. That third reader is why each clause in the template carries an evidence field.
Adopting the template does not mean your business meets any framework, standard or law, and it will not by itself carry a customer review or an audit. What it can do is make your practice clear enough for the person reviewing it to judge.
For what a policy is meant to achieve, the NIST Cybersecurity Framework 2.0 is a useful reference. It describes outcomes and does not prescribe how they should be achieved.
Its Govern outcomes include a policy for managing cybersecurity risks that is established, communicated and enforced (GV.PO-01), and that is reviewed and updated to reflect changes in requirements, threats, technology and the organisation's mission (GV.PO-02).
Keep only the clauses your team will actually follow
A clause nobody follows is worse than no clause at all.
Left in the policy, it reads as a control you run. It gets copied into a questionnaire answer, a reviewer asks for the record, and the distance between the document and the practice becomes the problem. A short policy that matches what happens holds up better than a long one that describes a team you do not have.
Before you adopt a clause, ask its owner one question: could you show the evidence for this today, from a system you already use? If yes, keep it. If the honest answer is "not yet, but we mean to", keep it and add an exception with an expiry date.
If the answer is "we are not going to do this", delete it, and if a customer later asks about that control, say so plainly.
Small teams tend to stumble on the same three things: access reviews nobody books, logs nobody reads, and supplier arrangements that were never written down. The template keeps each of them modest. An access review is an exported list of accounts with each one marked kept, reduced or removed.
The logging alerts clause names the few alerts that reach a real person instead of promising round-the-clock monitoring. The supplier clauses ask for a register and three things in each agreement, not a full assessment programme.
The seven modules and what each one decides
Each module stands on its own, so you can adopt the ones that apply and delete the rest.
Clauses are numbered by module letter (A1, C2, D3 and so on), which lets a questionnaire answer or a supplier agreement point to one clause instead of paraphrasing the policy.
- Access (A): who holds administrator rights to the application and the services around it, how access is granted, reviewed and removed, and where the application's own roles are written down.
- Change management (C): how code and configuration reach production, who looks at a change before release, where backend secrets live, and which changes count as security-relevant.
- Dependencies (D): the inventory of libraries, plugins and browser scripts, the update time frames you set by severity, and who receives advisory alerts.
- Logging (L): which logs exist, what they record and leave out, where they are kept and for how long, and which alerts reach a person.
- Incident escalation (I): who hears first, what nobody does before a copy is taken, and which outside specialist runs the response.
- Supplier access (S): which suppliers can reach the code, hosting or data, under what agreement, and what happens to their access when the work ends.
- Security testing (T): what prompts a test, how it is authorised and scoped, how findings are tracked to a retest result, and where outside security reports arrive.
Filling in the owner, evidence, exception and review fields
Every clause ends with owner, evidence and review fields, and some carry an exceptions line as well.
They turn a statement of intent into something a reviewer can check, and they are where most first drafts go wrong.
Owner
A named person, not a team or a job title nobody holds. The owner has the authority to change the practice and the time to keep it going. When an outside developer does the work, the owner is still someone inside the business who checks that the evidence exists; the developer can be named in the clause as the person who carries it out.
Evidence
Name the system, the record and where it is kept: "administrator list exported from the hosting control panel, saved to the Security folder after each review" rather than "access is reviewed". The strongest evidence is a record a system produces anyway, such as merge history in the code repository, a deployment log or the payment provider's user list. Evidence written up only for an audit tends to stop once the audit is over.
Exceptions
Departures from a clause go to the exception register at the end of the policy. The one exception is C1, where an emergency change is written up in a record of its own, with the reason. Keeping to one register means a gap is visible in one list rather than scattered through emails.
Review
A date or a set interval, the events that call for a review, or both, as the clause's Review line asks. Where a clause sets a number, such as the business days allowed to remove a leaver's access or the days allowed to apply a critical update, write the number you already meet. Tightening a number you meet is easier than explaining one you miss.
Framework references are pointers, not proof
Some clauses carry a "See" line naming a requirement in the OWASP Application Security Verification Standard (ASVS) 5.0 or an outcome in the NIST Cybersecurity Framework 2.0.
Each line names only the requirement, outcome or category its clause relates to, so a developer or reviewer can read the original wording. A reference does not mean the clause meets that requirement, and adopting the policy does not mean the application does.
Several ASVS 5.0 chapters open with documentation requirements, and a policy is a sensible place to say where that documentation lives. The secure coding and architecture chapter notes that many requirements for a secure and defensible architecture depend on clear documentation of decisions about how specific security controls are implemented and which components the application uses.
Requirement 15.1.1 asks that application documentation defines risk-based remediation time frames for third-party components with vulnerabilities and for updating libraries in general.
Requirement 16.1.1 asks for an inventory of the logging performed at each layer of the technology stack, and 8.1.1 asks that authorisation documentation defines the rules for function-level and data-specific access.
Because identifiers may change between versions, the ASVS says it is preferable for other documents to cite its requirements with the version in front, such as v5.0.0-1.2.5. The template follows that format, so each reference keeps pointing to the wording it was written against after the standard moves on.
OWASP says it does not certify vendors, verifiers or software, and that any assurance, trust mark or certification claiming compliance with the ASVS is not officially endorsed by OWASP. This template is not a certification against the ASVS, the NIST framework or anything else.
If a customer questionnaire asks for ISO 27001 certification or an Essential Eight maturity level, the template does not describe either, and the customer or auditor asking is the right person to say which document they mean.
Exceptions: recording the gap instead of hiding it
An exception is a dated decision that one clause does not apply in one place for a set period, with a named person accepting the risk and, where possible, something in place to reduce it meanwhile.
It keeps the policy honest without deleting the clause. When a reviewer asks about that control, the answer is already written: here is the gap, here is who accepted it, here is when it closes.
Four rules stop a register turning into a list of permanent excuses.
- Every exception has an expiry date. On that date it is closed because the clause is now met, or renewed with a new reason and a new date. Renewal without a reason is not an option.
- The person who accepts the risk has authority over the part of the business it falls on and, where the team is big enough, is not the person who asked for the exception.
- Each exception says what reduces the risk until it closes, such as a feature switched off or access narrowed. "Nothing" is an acceptable answer, as long as it is written down.
- Closed exceptions stay in the register with their closing date and the evidence that closed them, so anyone reading it can see that gaps are tracked to an end.
Why the incident and testing modules point outward
Two modules deliberately stop short of doing the work themselves.
Both record who decides and who is called, and leave the specialist work to the people who do it.
Incident escalation names the people, not the response
The incident module covers escalation: who hears about a suspected problem first, who makes the next call, and the few things nobody does before a copy is taken, such as deleting files or accounts or restoring a backup over the live site. The response itself belongs to your hosting provider and to an incident response provider or website cleanup specialist named in the clause, chosen before you need one.
Clever Ops does not provide incident response, forensic investigation, malware cleanup or monitoring, so the module refers that work out, as the hacked website guide in this library does.
Decisions about telling customers, an insurer, police or a regulator are taken with legal advice, and the clause names who takes them rather than describing what any law requires.
Testing is tied to change, at a frequency you set
The testing module does not set a calendar interval. It lists the events that should prompt a test or review: a new feature that handles payments, personal information or file uploads; a change to sign-in, roles or the way customer data is kept apart; a new integration or public API; a change of host, developer or agency; and a customer, contract or auditor asking for a recent test.
If a contract or auditor specifies a frequency, the clause records it with the document it came from. Otherwise the frequency is your decision, and the clause leaves room for it.
Whoever does the testing, the module asks for the same things: written authorisation for each target, a written scope, a report that says what was and was not tested, and a retest that marks each finding fixed, partially fixed, still reproducible or not retested. If the tester also built or hosts the application, the policy records that the work is a review by the builder rather than an independent test.
When Clever Ops built or hosts a system, we call the work a security review rather than an independent test, and the report says who built and who hosts it.
What a test report can and cannot evidence
A test report is good evidence for the testing module, and it can show how access rules and outdated components looked on the build that was tested. It says much less about logging. OWASP's Top 10:2025 guidance notes that testing whether logging and monitoring are in place, in use and effective can only be done with interviews and a sampling of effective incident responses, and that penetration testers may only be able to determine that they have invoked incident response in a test environment, which is rarely monitored in the same way as production. The evidence for the logging module comes from your own records.
Putting the policy into use
Most of the work is deciding rather than writing.
This order keeps the decisions with the people who will carry them out.
- Pick one application and write its name, environments, hosting and supplier list into the scope section. A second application gets its own copy.
- Delete the modules that do not apply. A site with no customer logins and no outside developer may need little more than access, dependencies, incident escalation and testing.
- Go through each remaining clause with its owner, fill in the numbers and the evidence location, and move anything not yet true into the exception register.
- Gather the first round of evidence before anyone signs. A clause whose evidence cannot be found on day one belongs in the register until it can.
- Have the policy approved by someone with authority over the business, keep it where your developers will see it, and give each supplier the supplier access module and any clause they carry out.
- Put the review date in a calendar, and list the events that bring it forward: a new supplier, an incident, a test report, or a change to how the application handles payments or personal information.
Frequently asked questions
The application and suppliers it covers, who owns it, and clauses for access, change management, dependencies, logging, incident escalation, supplier access and security testing. Each clause works best with an owner, the evidence that shows it is followed, a route for exceptions and a review date. Leave out anything your team will not do. An exception with an expiry date is more useful than a clause nobody follows.
No. A privacy policy is a public document about how you handle personal information, and what yours needs to say is a question for your lawyer. A website security policy is mostly internal: it records who can change the site, how changes are made and who is called when something goes wrong. A vulnerability disclosure policy is a third, separate document, published so that outside researchers know how to report a security problem to you.
No. It is not written against any of them, and adopting it says nothing about where your business stands under any framework or standard. A policy is one of the documents an auditor or customer may ask to see, alongside the records that show it is followed. Whether that is enough is their decision, so ask them what they expect before you rely on it.
Set a review date you will keep and write it on the policy, then bring the review forward when something changes: a new supplier, a new feature that handles payments or personal information, an incident, or a test report with findings. The NIST Cybersecurity Framework 2.0 treats review as an outcome in its own right: a policy that is reviewed, updated, communicated and enforced as requirements, threats, technology and mission change (GV.PO-02).
It can carry out clauses, such as reviewing changes or applying updates, and be named against them. The decisions about access, exceptions and suppliers stay with the business, so the policy owner should be someone inside it. Keep the business as the account owner of the domain, hosting, code repository and payment provider, so access does not leave with a supplier.
It is the public side of clause T7, which says where reports from outside researchers arrive and who triages them. The disclosure policy tells researchers how to send a report and what to expect afterwards. The NIST Cybersecurity Framework 2.0 lists established processes for receiving, analysing and responding to vulnerability disclosures as an outcome (ID.RA-08).
Yes, in a shorter form. Without customer logins or an outside developer, most of the access, change and supplier clauses come down to the hosting account, the domain registrar and the content management system's administrators. Keep the dependency clauses if the site runs plugins or themes, and keep the incident escalation module in full.
Scope the right test for your application
Tell us what you have built, who uses it and when it needs to launch. You will get a plain-English recommendation: a website audit, a web application test or a referral to a specialist.