ST4S assessment: how Safer Technologies 4 Schools works, and the evidence behind each security answer
For owners and technical leads of edtech businesses whose product has been asked to go through Safer Technologies 4 Schools (ST4S).
This guide sets out the process, the outcomes and the framework versions from the ST4S site and its Supplier Guide, read on 15 November 2026, shows which criteria a web application touches, and comes with a worksheet that labels every piece of evidence as a supplier record, a school decision or an independent test artefact.
In this guide
ST4S question-to-evidence worksheet for an edtech web application
A worksheet for an edtech supplier preparing ST4S answers about a web application. Its fifteen rows cover, among other areas, student data flows, sub processors, hosting, access control, incident handling and vulnerability management, using criteria from the ST4S Supplier Guide, release 2026.1, named by identifier and described in short form because the guide's copyright notice reserves its wording: read each criterion in the guide before you answer it.
How to use itHide the instructions
Each piece of evidence carries one of three labels. A supplier record is a document, setting, register or statement the supplier holds. A school decision is a choice the school or its education authority makes. An independent test artefact is a dated result from testing the running product by a tester outside the team that builds and runs it; when your own team runs a row, file the result as a supplier record. Everything named below is invented: an invented edtech app on its own staging copy, with two invented schools. Test School North has administrator adm-n, teachers tch-n1 and tch-n2, Year 5 students stu-n07 and stu-n08, and guardian gdn-n07, linked to stu-n07. Test School South has teacher tch-s1 and student stu-s03, whose records the North accounts try to reach. Every test account uses an email address and phone number the supplier controls, and no real student's, guardian's or staff member's record is used anywhere. Rows 3, 5, 11, 12, 14 and 15 are read-only. Rows 1, 2, 4, 6, 7, 8, 9, 10 and 13 send requests or sign in, and each carries its limits in its own fifth column. A service run by another business, for example a hosted sign-in or email provider, may be tested only within what its own terms permit: read them and ask the provider before sending anything, as rows 6, 8 and 9 repeat. The example answers are invented and show the shape of an answer that stays inside its evidence; write yours from your own evidence, because each answer is your attestation. A row's result is evidence about what it read or tested, never a finding that a criterion is met: the ST4S Team analyses and validates responses, and the school or education authority decides what to accept.
15 rows
Row, criteria and status in the guide
What the criteria ask, in short
1. Student data flows: P9 and P10, with the tier in Appendix A. Neither is hash-marked; both are asked in the Readiness Check.
- What the criteria ask, in short
- P9: which of the data types the guide lists the service would reasonably capture, store or process when used for its intended purpose; the list includes, among others, health and medical details, records of behaviour incidents, reason for absence, government related identifiers and location tracking data. P10: which kinds of functionality the service offers, each option marked with the tier it brings. Appendix A: the tier is the highest one the service qualifies for across data, functionality and reasonableness.
- Where an edtech web app meets them
- Sign-up and roster import forms, class lists, student profiles, free-text fields where teachers write comments, file uploads, and any analytics, error-reporting or support service the app sends data to.
- Evidence to hold, each item labelled
- Supplier record: a field-by-field data map taken from the build, listing each student field, why it is collected, where it is stored and every service that receives it. Independent test artefact: for one invented student, the outside hosts the browser sent student fields to on the build tested (file it as a supplier record if your own team runs it). School decision: whether the school seeks parental or student consent before it adds students, which the guide links to the Medium rating.
- How to produce it, and the row's limits
- Read-only first: list every student field from the forms, import templates and database schema, flagging free-text fields that can hold more than the app asks for. Then sign in as tch-n1 at Test School North, create invented student stu-n07 with every field filled with invented values, and have your developer record the host name of every outside request the browser makes and which invented values appear in each, with no key, token or cookie value written down. Sends requests or signs in, on the staging copy where one exists. Work only on your own product, with only the invented schools and accounts named in the description; no real school, student, guardian or staff member's record is used. Fix the number of requests before you start, counting each sign-in as one, keep it small and send no load. Check what your hosting provider's testing policy asks of you, such as notice or approval, before the first request. Point every email address and phone number on the test accounts at an inbox or handset you control, so anything the row sends reaches you. Leave passwords, keys and session tokens out of the record, along with any reset, invitation or download link that opens an account or file for whoever has it, and open such a link only for an invented account. If a response holds a real person's details, stop, note what appeared without copying it, and handle it through row 11.
- Invented example answer
- Invented example: "For Test School North, the app stores each student's name, year level, class, reading level and teacher comments. It asks for no health details, absence reasons or government identifiers. Teacher comments are free text, so our data map flags them as a field that can hold information we do not ask for. On the build tested, student fields reached our own API host only."
- What it cannot show, and who decides
- A browser capture sees what the browser sends, not transfers between servers, so the data map has to cover those. Appendix A is a self-assessment the supplier works through, and the ST4S Team analyses and validates the responses the tier rests on. Whether a field is needed is for the supplier to answer with its privacy adviser.
2. Sub processors and third-party providers: PR17 (hash-marked, on the Tier 1 and Tier 2 minimum criteria lists) and P12 (a Tier 1 question). Both are asked in the Readiness Check.
- What the criteria ask, in short
- PR17: whether the privacy policy or another published document describes all the service's sub processors, covering each one's name and organisation, contact information, the data types disclosed and the purpose, the lawful basis for processing where applicable, and the countries where data may be processed or stored. P12: whether, for every outside provider that forms part of the solution or provides services to the organisation, the supplier keeps a current inventory, assesses and manages their risks regularly, and has contracts requiring them to follow its security and privacy policies.
- Where an edtech web app meets them
- Hosting and database providers, email and SMS delivery, analytics, error reporting, support chat, video or AI services, and any script or font the app loads from another host.
- Evidence to hold, each item labelled
- Supplier record: the published sub processor list, the provider inventory and the contracts behind it. Independent test artefact: the outside hosts the browser contacted while invented users worked through the main features, compared with the list (file it as a supplier record if your own team runs it).
- How to produce it, and the row's limits
- Read-only first: compare the published list with your provider inventory, invoices and hosting accounts. Then, on staging, sign in as tch-n1 and as stu-n07 and use each main feature once while your developer records every outside host the browser contacts, noting host names only. Sends requests or signs in, on the staging copy where one exists. Work only on your own product, with only the invented schools and accounts named in the description; no real school, student, guardian or staff member's record is used. Fix the number of requests before you start, counting each sign-in as one, keep it small and send no load. Check what your hosting provider's testing policy asks of you, such as notice or approval, before the first request. Point every email address and phone number on the test accounts at an inbox or handset you control, so anything the row sends reaches you. Leave passwords, keys and session tokens out of the record, along with any reset, invitation or download link that opens an account or file for whoever has it, and open such a link only for an invented account. If a response holds a real person's details, stop, note what appeared without copying it, and handle it through row 11.
- Invented example answer
- Invented example: "Our published sub processor list names our cloud host, email delivery service and error-reporting service, with the data each receives, its purpose and where it processes data. On the build tested, the browser also loaded fonts from one further host that received no student fields, and we have added it to the list."
- What it cannot show, and who decides
- A browser capture cannot see a sub processor that only the server talks to, or what any provider does with the data it receives. Whether the list meets PR17 for your tier is for the ST4S Team.
3. Hosting and location: H1 and H5, both asked in the Readiness Check, H5 hash-marked and on both minimum criteria lists; and H2.
- What the criteria ask, in short
- H1: the countries where user data and any related data, such as metadata, logs and user content, is held or processed, across every part of the service: the live service, backups, disaster recovery, and test and development environments. H2: the countries from which the supplier's people, among them support, administration, development and testing staff, plus outside contractors or associates, can access or view that data, backups and recovery included. H5: whether customers would get advance notice if the cloud infrastructure, the data hosting environment, the supplier's staff, or anyone able to reach unencrypted customer data moved or expanded to another country.
- Where an edtech web app meets them
- The regions chosen for the app servers, database, file storage, backups and logs, the staging and development copies, and the locations of anyone who supports schools or administers the platform.
- Evidence to hold, each item labelled
- Supplier record: the region of each component as configured, the staff and contractor locations, and the clause in your school terms on advance notice of a move. School decision: whether the school, or its education authority, accepts the hosting location; section 2.2.4 of the guide gives a member accepting only data hosting within its jurisdiction as an example. Independent test artefact: none, because a test from outside cannot establish where data is stored.
- How to produce it, and the row's limits
- List each component and read its region in your providers' consoles or infrastructure files, then list where each person with access to student data works from. Read-only. Nothing is sent to the product: the row reads documents, settings, consoles and lists. Skip any screen or file that would display a password, key or secret.
- Invented example answer
- Invented example: "All live data, backups, logs and our staging copy for Test School North sit in our cloud provider's Sydney region, and support staff access student data from Australia only. Our school terms commit us to notify schools before any move of data or staff access to another country."
- What it cannot show, and who decides
- An address lookup, or where a server appears to be on the internet, does not show where data is stored or backed up. The response is the supplier's attestation, which the ST4S Team assesses, and acceptance of the location sits with the school or education authority.
4. Data in transit and certificates: S1 and S5. Both are hash-marked, on both minimum criteria lists and asked in the Readiness Check.
- What the criteria ask, in short
- S1: the minimum encryption protecting all data in transit, including between users, web applications and system components such as databases; the guide grades the options, placing TLS 1.1 or below in its weak option and TLS 1.3, with TLS 1.2 allowed for compatibility, in its strong option. S5: whether every web server of the service uses certificates from a reputable, trusted signing authority.
- Where an edtech web app meets them
- Every public hostname: the app, its API, file and image storage, the sign-in page and any marketing site that shares a login; plus the connections from the app to its database and other internal components.
- Evidence to hold, each item labelled
- Independent test artefact: for each public hostname, the protocol versions it accepted and the certificate issuer, from an external check on the date tested (file it as a supplier record if your own team runs it). Supplier record: the settings for internal connections, such as the database connection configuration.
- How to produce it, and the row's limits
- List every public hostname from your DNS records and app configuration, and check only hostnames your business owns. Connect to each one a handful of times to record the protocol versions and certificate it offers, and read the internal connection settings without opening any value that holds a key or secret. Sends a handful of plain requests, with no sign-in and no form submitted, to the public hostnames your business owns, because those hostnames are what this row reads. Work only on your own product, with only the invented schools and accounts named in the description; no real school, student, guardian or staff member's record is used. Fix the number of requests before you start, counting each sign-in as one, keep it small and send no load. Check what your hosting provider's testing policy asks of you, such as notice or approval, before the first request. Point every email address and phone number on the test accounts at an inbox or handset you control, so anything the row sends reaches you. Leave passwords, keys and session tokens out of the record, along with any reset, invitation or download link that opens an account or file for whoever has it, and open such a link only for an invented account. If a response holds a real person's details, stop, note what appeared without copying it, and handle it through row 11.
- Invented example answer
- Invented example: "On the date tested, every public hostname of the app, its API and its file storage accepted TLS 1.2 and TLS 1.3 only and presented a certificate from a public certificate authority. The app connects to its database over an encrypted connection, as set in the attached configuration."
- What it cannot show, and who decides
- An outside check cannot see connections between internal components, or how stored data is encrypted, which is criterion S2. Which option meets the minimum for your tier is set in the guide and judged by the ST4S Team.
5. Production data in test environments: S13. Hash-marked, on both minimum criteria lists and asked in the Readiness Check.
- What the criteria ask, in short
- S13: whether production data is used in non-production environments such as test and development; its failing option is production data used without identical security controls applied and without de-identification.
- Where an edtech web app meets them
- Staging and development copies of the app, demonstration environments shown to schools, and any copy a developer restores to their own machine.
- Evidence to hold, each item labelled
- Supplier record: how staging is filled (a seeding script or procedure), who can reach staging, and a dated statement that no production records are copied into it. Independent test artefact: the scope section of a test report naming the environment and the invented accounts used.
- How to produce it, and the row's limits
- Read the seeding script or procedure and the staging access list, and check the list of schools in the staging database for any name that is not one of your invented ones. Read-only. Nothing is sent to the product: the row reads documents, settings, consoles and lists. Skip any screen or file that would display a password, key or secret.
- Invented example answer
- Invented example: "Staging holds two invented schools, Test School North and Test School South, created by a seeding script. No production records are copied into staging, and our last penetration test ran there with invented accounts only."
- What it cannot show, and who decides
- A test report shows which environment one test used, not how every copy is filled. The ST4S Team assesses the response.
6. Unique sign-in and access denied by default: A1 and A13. Both are hash-marked, on both minimum criteria lists and asked in the Readiness Check.
- What the criteria ask, in short
- A1: whether every user of the service, such as teachers, parents, students and administrators, is uniquely authenticated and identifiable; an option that excludes students, with the reason and the controls for students detailed, is also marked acceptable for both tiers. A13: the default approach to user access, where protection by default (deny unless approved) is the acceptable option and protection by exception (allow unless specifically denied) is the failing one.
- Where an edtech web app meets them
- Sign-in for each role, class join codes or shared logins, teacher and administrator pages, and the API routes behind them.
- Evidence to hold, each item labelled
- Supplier record: the role permission matrix, written before testing, and any student sign-in exception with its reason. School decision: which roles the school gives its staff, students and guardians. Independent test artefact: for each teacher and administrator page and API route tried, the response to a signed-out browser and to the stu-n07 and gdn-n07 sessions (file it as a supplier record if your own team runs it).
- How to produce it, and the row's limits
- Write the permission matrix for adm-n, tch-n1, stu-n07 and gdn-n07. Then, with no session, as stu-n07 and as gdn-n07, request each teacher and administrator page and API route by its address, and record the response to each. Sends requests or signs in, on the staging copy where one exists. Work only on your own product, with only the invented schools and accounts named in the description; no real school, student, guardian or staff member's record is used. Fix the number of requests before you start, counting each sign-in as one, keep it small and send no load. Check what your hosting provider's testing policy asks of you, such as notice or approval, before the first request. Point every email address and phone number on the test accounts at an inbox or handset you control, so anything the row sends reaches you. Leave passwords, keys and session tokens out of the record, along with any reset, invitation or download link that opens an account or file for whoever has it, and open such a link only for an invented account. If a response holds a real person's details, stop, note what appeared without copying it, and handle it through row 11. Where a step passes through a service run by another business, for example a hosted sign-in, email or SMS provider, read that provider's terms on testing and ask it before sending anything; this worksheet says nothing about what any provider allows.
- Invented example answer
- Invented example: "Every teacher, student, guardian and administrator at Test School North signs in to their own account. On the build tested, a signed-out browser, student stu-n07 and guardian gdn-n07 were each refused every one of the 14 teacher and administrator routes tried."
- What it cannot show, and who decides
- A result covers the routes and accounts tried on that build, not routes nobody listed. The IDOR and broken access control guide sets out the underlying check. The ST4S Team assesses the response.
7. Separation between schools: S4 and PR14. Both are hash-marked, on both minimum criteria lists and asked in the Readiness Check.
- What the criteria ask, in short
- S4: for a multi-tenant service, which controls keep each customer's data apart: unique customer identifiers in shared tables, separate tables or databases, or dedicated instances, environments or virtual private clouds; none of these is the failing option. PR14: whether any discovery feature lets users at one school find, reach or discover users or personal information at another school or organisation, and which restrictions apply; discovery with none of the listed controls is the failing option.
- Where an edtech web app meets them
- Every list, search, report, export and record page a school account can open, and any feature that finds people or shares data across schools, such as a student transfer.
- Evidence to hold, each item labelled
- Supplier record: how tenancy is separated in the design. School decision: whether the school switches on any discovery or sharing feature, where the product leaves that to it. Independent test artefact: the response to each cross-school request from each role (file it as a supplier record if your own team runs it).
- How to produce it, and the row's limits
- As adm-n, tch-n1 and stu-n07 of Test School North, open tch-s1's class, stu-s03's record and a Test School South report by address, run each search and list with the school or class value changed to South's, and use any discovery feature to look for stu-s03. Sends requests or signs in, on the staging copy where one exists. Work only on your own product, with only the invented schools and accounts named in the description; no real school, student, guardian or staff member's record is used. Fix the number of requests before you start, counting each sign-in as one, keep it small and send no load. Check what your hosting provider's testing policy asks of you, such as notice or approval, before the first request. Point every email address and phone number on the test accounts at an inbox or handset you control, so anything the row sends reaches you. Leave passwords, keys and session tokens out of the record, along with any reset, invitation or download link that opens an account or file for whoever has it, and open such a link only for an invented account. If a response holds a real person's details, stop, note what appeared without copying it, and handle it through row 11.
- Invented example answer
- Invented example: "Each school's records carry a school identifier that every query checks. On the build tested, all 23 requests from Test School North accounts for Test School South records were refused, each answered exactly as a request for a missing record is, no search returned a South name, and the app has no feature for finding users at another school."
- What it cannot show, and who decides
- Two invented schools with a handful of records show the boundary on the routes tried, not on routes nobody listed or under load. The SaaS security checklist covers tenant isolation in more depth. The ST4S Team assesses the response.
8. Passwords and password reset: A2 (hash-marked, on both minimum criteria lists) and A11 (hash-marked, on the Tier 1 list). Both are asked in the Readiness Check.
- What the criteria ask, in short
- A2: whether every password is hashed with a strong, salted, one-way algorithm in line with OWASP recommendations; the service accepts passphrases of 64 characters or more and Unicode characters, with a minimum length set by relevant standards; each new password is compared with lists of weak, common or breached ones; and sign-in attempts are throttled. A11: whether a reset issues a single-use, time-limited link or token, sends it only through verified, trusted channels, makes the user choose a new password as soon as the token is validated and voids the token once used or expired, rate-limits and monitors reset requests without revealing whether an account exists, and ends active sessions and refresh tokens after a successful reset.
- Where an edtech web app meets them
- The sign-up, sign-in, change-password and reset pages and their API routes, and the reset email.
- Evidence to hold, each item labelled
- Supplier record: the hashing algorithm and its settings, from the code or configuration. Independent test artefact: how sign-in and reset behaved for test accounts on the build tested, request by request (file it as a supplier record if your own team runs it).
- How to produce it, and the row's limits
- Read the hashing settings without opening any value that holds a key or secret. Then, with tch-n2's test account only: request a reset and use the link twice; try a link after the expiry the app states; request a reset for an address with no account and compare the response; sign in on a second browser, reset the password, and reload the second session; and send a small number of wrong passwords, fixed before you start, to see whether attempts are throttled. Lockout and throttling checks run on test accounts only. Sends requests or signs in, on the staging copy where one exists. Work only on your own product, with only the invented schools and accounts named in the description; no real school, student, guardian or staff member's record is used. Fix the number of requests before you start, counting each sign-in as one, keep it small and send no load. Check what your hosting provider's testing policy asks of you, such as notice or approval, before the first request. Point every email address and phone number on the test accounts at an inbox or handset you control, so anything the row sends reaches you. Leave passwords, keys and session tokens out of the record, along with any reset, invitation or download link that opens an account or file for whoever has it, and open such a link only for an invented account. If a response holds a real person's details, stop, note what appeared without copying it, and handle it through row 11. Where a step passes through a service run by another business, for example a hosted sign-in, email or SMS provider, read that provider's terms on testing and ask it before sending anything; this worksheet says nothing about what any provider allows.
- Invented example answer
- Invented example: "Passwords are hashed with Argon2id, with the settings in the attached configuration. On the build tested, the reset link for tch-n2 worked once and was refused on reuse and after 30 minutes, the reset page gave the same message for an unknown address, the second browser's session ended after the reset, and sign-in slowed after five wrong passwords."
- What it cannot show, and who decides
- A test sees behaviour, not how passwords are stored, so the hashing answer stays a supplier record. Among its examples of where criteria can apply differently in particular circumstances, section 2.2.5 of the guide lists password rules for students and young people, such as shorter passwords and other ways in. It says an alternative finding is not guaranteed and is reached in consultation with Working Group members; whether one is made is for the ST4S Team. The broken authentication guide covers these flows.
9. Multi-factor sign-in for school users: A4 (a Tier 1 question whose "No" option carries a hash) and A19. Neither is in the Readiness Check list.
- What the criteria ask, in short
- A4: whether the service offers multi-factor authentication to the people who use it, among them school staff, parents and students; the options run from offering it for all account types to mandating it for all non-student accounts or for all account types. A19: which secure authentication settings an administrator can apply to their school or users, such as a minimum password rule or enforcing multi-factor authentication.
- Where an edtech web app meets them
- Staff and guardian sign-in, the second-factor enrolment and challenge pages, the administrator settings page, and every other way into an account, such as password reset or a school single sign-on.
- Evidence to hold, each item labelled
- Supplier record: which multi-factor options the product offers and what an administrator can enforce. School decision: whether the school's administrator enforces multi-factor sign-in for its staff, where the product leaves that to the school. Independent test artefact: whether, once enforced for a test staff account, the second factor was asked for on every route into that account that was tried (file it as a supplier record if your own team runs it).
- How to produce it, and the row's limits
- As adm-n, enforce multi-factor sign-in for staff. Then, as tch-n1, try to reach a teacher page by signing in normally, straight after a password reset, and through any other sign-in route the product offers, without completing the second factor. Sends requests or signs in, on the staging copy where one exists. Work only on your own product, with only the invented schools and accounts named in the description; no real school, student, guardian or staff member's record is used. Fix the number of requests before you start, counting each sign-in as one, keep it small and send no load. Check what your hosting provider's testing policy asks of you, such as notice or approval, before the first request. Point every email address and phone number on the test accounts at an inbox or handset you control, so anything the row sends reaches you. Leave passwords, keys and session tokens out of the record, along with any reset, invitation or download link that opens an account or file for whoever has it, and open such a link only for an invented account. If a response holds a real person's details, stop, note what appeared without copying it, and handle it through row 11. Where a step passes through a service run by another business, for example a hosted sign-in, email or SMS provider, read that provider's terms on testing and ask it before sending anything; this worksheet says nothing about what any provider allows.
- Invented example answer
- Invented example: "Administrators can require multi-factor sign-in for all staff, and Test School North has chosen to. On the build tested, tch-n1 could not reach any teacher page without the second factor, including straight after a password reset."
- What it cannot show, and who decides
- Two-factor authentication for the supplier's own people, covering staff and contractors who reach its systems remotely, its system administrators and support staff, and privileged accounts, is a separate criterion, A5, answered from the supplier's records. The MFA security guide covers the ways around a second factor. The ST4S Team assesses the response.
10. Deletion, export and backups: PR13 and D3 (both hash-marked, on both minimum criteria lists, asked in the Readiness Check), D5 and D6.
- What the criteria ask, in short
- PR13: whether the organisation has a process through which customers can ask for access to, correction of or deletion of all personal information about them; the options are graded by cost and by whether requests are resolved within 3 months. D5: a process to delete inactive or dormant users and their data. D6: whether all user data can be exported in a reusable form. D3: whether a full restore from backup is tested when backups are first set up, again whenever major infrastructure changes, and at least once a year.
- Where an edtech web app meets them
- The delete-student and delete-account features, the school export, the route a school uses to ask for deletion, search and lists after a deletion, file links to the student's uploads, and backups.
- Evidence to hold, each item labelled
- Supplier record: the request procedure and its timeframe, the dormant-user rule, the export format, the backup retention period and the last restore test record. School decision: when the school asks for a student's records to be deleted or exported. Independent test artefact: whether an invented student deleted through the product's own feature still appeared anywhere a school account could reach, on the build tested (file it as a supplier record if your own team runs it).
- How to produce it, and the row's limits
- Create invented student stu-n08 with an invented upload. As adm-n, export Test School North's data and check the file can be reused, then delete stu-n08 through the product's own feature and look for the student in screens, search, lists, a fresh export and the old file link. Delete nothing except records invented for this row, and only read the restore test record. Sends requests or signs in, on the staging copy where one exists. Work only on your own product, with only the invented schools and accounts named in the description; no real school, student, guardian or staff member's record is used. Fix the number of requests before you start, counting each sign-in as one, keep it small and send no load. Check what your hosting provider's testing policy asks of you, such as notice or approval, before the first request. Point every email address and phone number on the test accounts at an inbox or handset you control, so anything the row sends reaches you. Leave passwords, keys and session tokens out of the record, along with any reset, invitation or download link that opens an account or file for whoever has it, and open such a link only for an invented account. If a response holds a real person's details, stop, note what appeared without copying it, and handle it through row 11.
- Invented example answer
- Invented example: "When Test School North asked us to delete invented student stu-n08, the record and the student's upload disappeared from every screen, search, export and file link a school account could reach on the build tested. Backups keep it until they expire 90 days later, as our terms say. Our last full restore test ran after the database upgrade in term 2."
- What it cannot show, and who decides
- A test cannot see inside backups or confirm that a restore test happened, so those stay supplier records. Whether a timeframe or a cost meets PR13 for your tier is for the ST4S Team, and what privacy law says about deletion is a question for your adviser.
11. Incident handling: T6, T6A and T7 (hash-marked, on both minimum criteria lists, asked in the Readiness Check, with T6A asked only where T6 is answered yes) and I1.
- What the criteria ask, in short
- T6: whether a formal, written incident response plan is in place and in use, setting out who does what and requiring every security, privacy and online safety incident to be investigated, fixed and entered in a register that holds, at a minimum, when it happened, when it was found, what happened, what was done and who it was reported to. T6A: how often the plan is tested and updated. T7: whether affected customers or organisations hear as soon as possible once a data breach comes to light, with all the relevant details, such as who was affected and what was disclosed. I1: whether the organisation, platform or service has had a security or privacy incident or breach within three years, and if so whether each was handled under its incident response or privacy breach plan, notified to central agencies and the people affected, and investigated, with controls improved as a result.
- Where an edtech web app meets them
- Who in the business takes a report from a school, how a finding from a test or a school is logged, how schools are told, and the contact details schools are given.
- Evidence to hold, each item labelled
- Supplier record: the plan (EV9 in the guide's evidence list), the register, and the date and notes of the last test of the plan. School decision: what the school tells its own families once it has been notified. Independent test artefact: none for the plan itself; a test can surface an exposure, which then runs through the plan.
- How to produce it, and the row's limits
- Read the plan against the T6 list and check that the register holds each field T6 names. Find the date of the last exercise and what changed after it. Confirm that a test result showing a real person's details is logged as an incident and handled under the plan. Read-only. Nothing is sent to the product: the row reads documents, settings, consoles and lists. Skip any screen or file that would display a password, key or secret.
- Invented example answer
- Invented example: "Our incident response plan names our technical lead as incident owner and our privacy officer as the person who decides on notification, and its register holds each field T6 names. We tested the plan in an exercise built on an invented exposure of Test School North records and updated the contact list afterwards."
- What it cannot show, and who decides
- A penetration test does not evidence how a business responds to an incident. Whether a particular breach has to be reported under a law is for the supplier and its adviser, and the ST4S Team assesses the response to the criteria.
12. Vulnerability management: T1 (hash-marked, on both minimum criteria lists, asked in the Readiness Check), with evidence items EV10 and EV11.
- What the criteria ask, in short
- T1: a continuous monitoring plan in use across all of the supplier's systems and infrastructure, with automatic asset discovery, regular automated scanning for missing patches against a current vulnerability database, a penetration test after a major change or at least once a year, assessment of each vulnerability's likely impact and of fixes weighed on effectiveness, cost and existing controls, and fixes prioritised by risk. Its options distinguish meeting all of this from also using outside, independent testers for the penetration testing. EV10: a redacted copy of the service's most recent penetration test report. EV11: redacted copies of its most recent vulnerability assessment reports.
- Where an edtech web app meets them
- The app, its API, its hosting and every other system in the plan's scope, with the dates of major changes such as a new module or a hosting move.
- Evidence to hold, each item labelled
- Supplier record: the plan, the asset inventory, scan schedules and results, and a remediation log with priorities. Independent test artefact: the most recent penetration test report for the service, naming the build, dates, accounts and exclusions, with each finding's retest status, redacted for lodging.
- How to produce it, and the row's limits
- Check that the report covers the service and version the questionnaire names, and that its date falls after the last major change. Mark any finding still open and the date its fix is planned. Before commissioning a test, ask the ST4S Team what it will accept as EV10, including any tester certification it expects. Read-only. Nothing is sent to the product: the row reads documents, settings, consoles and lists. Skip any screen or file that would display a password, key or secret.
- Invented example answer
- Invented example: "Our continuous monitoring plan, attached, covers asset discovery, weekly automated scans, a penetration test at least annually and after any major change, and risk-based prioritisation. The most recent penetration test of the app and its API was run by an outside provider on the build released before term 3, and the attached redacted report shows two findings, both marked fixed after the retest."
- What it cannot show, and who decides
- One report shows one test of one build, not a plan, a scan programme or that later releases were tested. Section 6.8 says the ST4S Team may check documents with their author, for example by asking a penetration tester for proof of certification, and the Team decides whether a report is accepted.
13. Patching of internet-facing services: T3, a Tier 1 question, hash-marked, on the Tier 1 list and asked in the Readiness Check.
- What the criteria ask, in short
- T3: whether security fixes, updates or vendor mitigations go in no later than two weeks after release, or 48 hours where an exploit exists, for internet-facing services and their operating systems, the operating systems of workstations, servers, network devices and other ICT equipment, and device drivers and firmware.
- Where an edtech web app meets them
- Web servers, the frameworks and libraries the app runs on, the operating systems under them, and any software whose name or version shows from outside, such as in a response header.
- Evidence to hold, each item labelled
- Supplier record: the patch log with release and applied dates; EV12 in the guide's evidence list covers patch management standards or process. Independent test artefact: the software names and versions visible from outside on each public hostname on the date tested (file it as a supplier record if your own team runs it).
- How to produce it, and the row's limits
- Read the patch log for the last quarter. Then request the home page and one API route on each public hostname your business owns, and record any software name or version the responses reveal. Sends a handful of plain requests, with no sign-in and no form submitted, to the public hostnames your business owns, because those hostnames are what this row reads. Work only on your own product, with only the invented schools and accounts named in the description; no real school, student, guardian or staff member's record is used. Fix the number of requests before you start, counting each sign-in as one, keep it small and send no load. Check what your hosting provider's testing policy asks of you, such as notice or approval, before the first request. Point every email address and phone number on the test accounts at an inbox or handset you control, so anything the row sends reaches you. Leave passwords, keys and session tokens out of the record, along with any reset, invitation or download link that opens an account or file for whoever has it, and open such a link only for an invented account. If a response holds a real person's details, stop, note what appeared without copying it, and handle it through row 11.
- Invented example answer
- Invented example: "Our patch log records the release and applied dates of every security update to internet-facing components, and the attached extract covers the last quarter. On the date tested, no software version was visible from outside on any public hostname."
- What it cannot show, and who decides
- Versions visible from outside are a small part of what T3 covers, and a date in a log is the supplier's record. The ST4S Team assesses the response.
14. Security testing in development: Q5. Hash-marked, on both minimum criteria lists and asked in the Readiness Check.
- What the criteria ask, in short
- Q5: whether every application development for the service is assessed under a security testing method that follows the guidance in what the criterion calls the "latest" industry standard frameworks, giving OWASP Testing Guide v4.2 and BSIMM as examples; the options separate testing that partially satisfies that guidance from testing that fully satisfies it, and ask the supplier to name the framework.
- Where an edtech web app meets them
- Each release of the app and its API, the release checklist, and code review and test records.
- Evidence to hold, each item labelled
- Supplier record: the secure software development lifecycle process (EV13 in the guide's evidence list) and per-release test records naming the framework. Independent test artefact: a penetration test report whose method section names the framework the test referenced.
- How to produce it, and the row's limits
- Read the release process and pick two recent releases; confirm each has a security test record that names its framework. Read-only. Nothing is sent to the product: the row reads documents, settings, consoles and lists. Skip any screen or file that would display a password, key or secret.
- Invented example answer
- Invented example: "Each release is checked against a list drawn from the OWASP Testing Guide v4.2, recorded in the release ticket. Our outside penetration test report names the same guide in its method section."
- What it cannot show, and who decides
- A report that names a framework shows how one test was organised, not that every development is assessed, and it does not certify the product against that framework. The ST4S Team assesses the response.
15. Integrations and school accounts in other services: INT7 and A16B (both hash-marked, on both minimum criteria lists) and A16A, all asked in the Readiness Check.
- What the criteria ask, in short
- INT7: whether, for native and partner-built integrations, agreements with every connected third-party service cover the purpose of the data sharing, the scope and scale of the data shared, the security and privacy controls in recipient systems and ownership of the data, and are updated when those change. A16A: whether the supplier or its service asks a school to hand over login details for another system or service, or to set up a user account in one. A16B: where accounts are created in third-party services, whether enforceable written agreements covering that arrangement are in place with all of them.
- Where an edtech web app meets them
- Roster and timetable syncs, single sign-on, learning management connections, partner integrations, and any setup step that asks a school for a login to another system.
- Evidence to hold, each item labelled
- Supplier record: the list of integrations with the agreement behind each, and any setup step that asks schools for credentials or accounts. School decision: which integrations the school connects, and which accounts it creates in other systems. Independent test artefact: none for the agreements; testing the routes an integration calls is covered in the API security guide.
- How to produce it, and the row's limits
- List every integration a school can connect, the data it moves and the agreement behind it, and note any setup step that asks a school for credentials to another system. Read-only. Nothing is sent to the product: the row reads documents, settings, consoles and lists. Skip any screen or file that would display a password, key or secret.
- Invented example answer
- Invented example: "Schools can connect our app to two roster services, each under an agreement covering purpose, scope, scale, recipient controls and data ownership. Setup never asks a school for its login to another system."
- What it cannot show, and who decides
- Whether an agreement is adequate for INT7 is assessed by the ST4S Team, and what a contract means in law is for the supplier's solicitor.
What an ST4S assessment is, and who runs it
The Safer Technologies 4 Schools (ST4S) home page presents the initiative as one standard way of evaluating the digital products and services that schools in Australia and New Zealand use, measured against a security and privacy control framework that the page calls nationally consistent.
The same page says ST4S was developed with all of Australia's state, territory, Catholic and independent school sectors, together with New Zealand's Ministry of Education. An ST4S assessment applies that framework to one supplier's product: the supplier answers a questionnaire, uploads documents where its answers call for them, and the ST4S Team analyses and validates what was submitted.
Education Services Australia (ESA) administers ST4S for the state and territory governments, the Catholic and independent school sectors and New Zealand's Ministry of Education, according to the ST4S general information page, and the ESA team that co-ordinates assessments is its National Schools Interoperability Program (NSIP).
The ST4S page on its Working Group describes ESA as a not-for-profit organisation owned by Australian Education Ministers and says the ST4S team are ESA employees. The Working Group oversees the initiative, draws its members from government agencies and the K-12 sector in both countries, and meets monthly; its work at those meetings includes nominating products for assessment, putting forward and approving changes to the ST4S Framework, discussing privacy, security and related issues, and other governance work.
The members that page lists, as read on 15 November 2026, are the government education department or directorate of every Australian state and territory, the Commonwealth Department of Education, the New Zealand Ministry of Education, and Catholic and independent sector bodies. It says members may keep information about ST4S on their own websites, including official statements and policies.
This guide says nothing about what any one of them asks of a supplier: that is in the member's own pages, or in its request to you.
This page is written for the owner or technical lead of an edtech business. It describes ST4S from the ST4S site and from the ST4S Supplier Guide, release 2026.1, which the site's menus call the vendor guide. Nothing here is legal advice or an ST4S assessment, and nothing on this page, in its worksheet or in a test report produces an ST4S outcome.
What counts as evidence for your product is for the ST4S Team, and what a school or education authority accepts is for that school or authority. The Readiness Check page and the guide both say taking part in ST4S is voluntary; whether a particular customer has made it a condition of buying or using your product is something to read in your agreement with that customer and take to your own adviser.
Who can ask for an ST4S assessment, and how a supplier takes part
The general information page gives two routes in: the supplier nominates itself, or a school or education authority invites it.
The Readiness Check page names where a request may come from: Education Services Australia, a State or Territory Department of Education, a school, or a school authority (Catholic or Independent) that is a member of the ST4S Working Group. That list is about requests to take the Readiness Check, which the general information page puts first on either route.
Section 2.2 of the Supplier Guide adds that a supplier may in some circumstances be invited straight to the full assessment, for example where a Working Group member has asked for it.
The Readiness Check
The Readiness Check is a tool suppliers use to assess themselves. Its page says it suits products and services used in a K-12 setting that process or handle personally identifiable or sensitive information and operate online, wholly or in part, and that a service with no online component is not suited to it.
It has two steps: a short survey of critically important criteria from the vendor guide, which returns feedback on readiness, and a second, optional step that puts the service forward to be considered for a full assessment. A supplier may take it more than once.
Initial results stay private unless the supplier formally submits them, and the page says suppliers cannot use Readiness Check results to promote their services with current or future customers. The page also says participation in ST4S is voluntary.
From nomination to a school-level report
Section 2 of the Supplier Guide says that once a service is put forward for full assessment, the ST4S Team checks it and the Working Group sets priorities each month within a monthly cap, with members nominating services according to school demand and other factors. The Readiness Check page says its results do not guarantee a priority invitation to the full assessment, or compliance under it.
An invited supplier completes the full questionnaire and lodges documentation. Section 2.2.1 sets out four stages: submission; a precheck and evidence review, in which the Team compares the submission with the minimum criteria, looks at the supplier's website, runs technical checks and goes through the documents and evidence supplied; assessment and report creation, during which the supplier is given a chance to clarify its responses and may seek to remediate further, with any extra time subject to the Team's review; and finalisation.
Section 5.4 says a supplier should allow at least 3 months after it submits the Full Assessment questionnaire.
What the supplier attests to
Section 5.3 of the guide says that when it submits the questionnaire, a supplier must confirm that what it has provided is true, correct, accurate and up to date, and in no way misleading, and must agree to back its responses with further information or evidence on request.
Section 5.1 adds that the person completing the Readiness Check or Full Assessment on the supplier's behalf must be duly authorised and must have express written permission from the chief executive, the chief information officer or another senior officer with comparable authority over executive decisions and the company's strategy. Each answer is therefore the supplier's own statement, whoever helped gather the evidence behind it.
What the outcome is called, who sees it, and what it is not
Section 2.2.3 of the Supplier Guide says the possible outcomes depend on the product's tier.
For Tier 1 and Tier 2 services they are Low risk, Medium risk, High risk and Non-compliant. The overall outcome is the highest level of risk still left once every available treatment is in place, and a service is given Non-compliant when it fails a mandatory minimum standard.
Tier 3 services receive Use Responsibly or Use with Caution, and other outcomes include discontinued and non-participating. Appendix A, which the guide calls a tier self-assessment, sets the tier as the highest one a service qualifies for across three factors: the data it stores or processes, its functionality, and reasonableness, meaning its display of advertising or other material that may cause offence.
Section 2.2.4 says Medium, High and Use with Caution outcomes are meant to point school users to treatments they need to review and put in place when they use the product, such as configuration or reviewing logs.
It also says Medium is the rating products most commonly receive, and that this reflects how much it matters to schools to seek parental or student consent before personal information is disclosed or shared with the service.
Once the review is complete, the general information page says the final results go to the Working Group's members, the chief information officers of the education jurisdictions and New Zealand's Ministry of Education, and approved ST4S stakeholders may pass them on to their schools. Section 3.2 of the guide adds that information from an assessment may also be shared with other parties the Working Group or NEDAG nominates, and that with their approval results may go to other parties without notice to the supplier.
The Readiness Check page calls the use of ST4S results and reports from the full assessment a local decision.
Except where the outcome is non-compliant, the assessment was discontinued or no report was produced, section 2.2.2 says the supplier's nominated contact receives the final school-level report. Section 3.5 then sets conditions on its use, among them: the supplier may tell customers that it took part and meets ST4S's minimum required standards, naming the version of the standards and the version of its product assessed, but it must not pass the report or any extract to anyone outside its own organisation, publish or promote its specific rating, or claim the assessment covers other products or versions.
What an ST4S outcome is not
Section 5.1 says school-level reports are not an endorsement of, approval of or recommendation about using the product, and give no advice on its quality or licensing or on whether to buy or use it. Under section 3.5, a supplier must tell customers that an outcome does not necessarily mean it meets the local requirements of a State, Territory, country or non-government sector, and must not suggest, openly or by implication, that ST4S endorses, recommends or approves the product, or that it vouches for the product being fit for purpose.
Section 5.7 bars a supplier from suggesting it conforms to ST4S standards and criteria while the only step it has finished is the Readiness Check and it holds no recent, successful full assessment of the service that is still valid, and from making its own badge, images or content in the style of ST4S; one of its examples is the words "ST4S Certified" next to a tick.
Under section 2.2.4, an education jurisdiction, another ST4S member or NEDAG, the National Education Digital Advisory Group, may at its discretion accept an outcome or reject it, add requirements of its own, or carry out its own assessment activities.
The badge, and how long an outcome lasts
The ST4S Product Badge Program has its own eligibility criteria. Its page lists what a supplier must meet to be eligible: a product successfully assessed under a 2020+ version of the framework, with an outcome of Low risk, Medium risk or Use Responsibly; no advertising in the product, and no use or sharing of information, including de-identified information, for advertising, marketing, promotional or similar purposes, although a supplier may request an exemption, reviewed by the Working Group, where every condition the page lists is met; any other criteria ESA sets; and a completed licence agreement.
The page says eligible suppliers may be invited to take part once an assessment is satisfactorily completed, that the right to use the badge depends on complying with the licence agreement's terms, and that to keep using it a supplier must commit to a Readiness Check or a full assessment at least annually.
The ST4S page on validity says an assessment is valid for two years from the date of completion, and that products are due for reassessment every two years, or sooner if significant changes occur that could affect compliance with current privacy, security or online safety standards. Telling ST4S about such changes within the two years is, on that page, the supplier's responsibility.
Section 2.2.6 of the guide gives the same two years, unless the assessment is withdrawn before then, and section 3.5 says the ST4S Assessment Team may rescind, withdraw or modify an outcome.
Which framework version applies to your assessment
The vendor guide page ties each assessment to one numbered release of the ST4S framework, written as year and release number; in its own example, 2023.1 is the first release for 2023.
The framework versions page says the framework is usually updated twice a year, that a version marked Latest or Active is considered by the Working Group, and that a version marked Transitioning may be considered depending on the school or Working Group member.
Read on 15 November 2026, the versions list on that page labels v2026.1 "LATEST", v2025.1 and v2024.1 "CURRENT", v2023.2 and v2023.1 "TRANSITIONING" and v2022.2 "DEPRECATED". A paragraph on the same page, headed "What's current", says the latest version of the framework is v2025.1 and that versions 2024.1 and V2023.1 remain current. The two statements differ, so ask whoever sent the request which version your assessment will use.
The v2026.1 page gives 21 July 2026 as the original framework publication date and says 41 assessment questions were updated across the Full Assessment and the AI Module, with changes that include greater emphasis on child safety screening of vendor personnel where applicable and revised risk ratings for authentication controls.
The Supplier Guide for release 2026.1, document version V1.0, gives 21 June 2026 as the date of the version on its cover page, and 19 June 2026 against V1.0 in its version control table.
Section 2.2.6 of the guide says a reassessment will generally be necessary after changes such as a new owner of the service; new, changed or removed features that bring new categories into the assessment; many data types added or removed, or any change to data types ST4S classes as sensitive or protected; a change of hosting location or infrastructure; privacy policy changes that warrant a new assessment, such as introducing or changing advertising; and adding or removing AI features, or changing how data on the service may be used to train or develop AI models.
A release plan that includes any of these is worth raising with the ST4S Team before it ships.
ST4S security requirements a web application touches
Section 6 of the Supplier Guide holds the criteria.
It groups them as company and product details, Security (product function, hosting and location, technical, logging, access, HR, processes and testing, plans and quality, incidents, data deletion and retention, compliance controls and governance), Privacy, Interoperability, Safety, desktop and mobile applications and browser extensions, Evidence, and an AI Module.
Each criterion has an identifier, such as T1 or PR17, with response options and the tiers it applies to. The guide marks a minimum acceptable response, which can differ by tier, and puts a hash (#) on questions of high importance: failing the minimum on one of them results in a Non-compliant outcome, although the guide notes that some hash-marked items on its minimum criteria lists may instead result in a high risk outcome.
The guide is offered for download on the ST4S vendor guide page, under its current version. Its copyright notice says it must not be reproduced or published, in whole or in part, without ESA's prior written permission, so the worksheet on this page names each criterion by its identifier and describes it in short form.
Read the criterion itself in the guide before you answer it. Section 2.1 of the guide lists the criteria the Readiness Check presents, and each worksheet row says whether its criteria are among them.
This page sorts the criteria a web application touches into three groups.
- Criteria where a test of the running product can observe part of the answer, for the build and accounts tested: encryption in transit and certificates (S1, S5), whether every user signs in to their own account and access is denied unless granted (A1, A13), how passwords and resets behave (A2, A11), multi-factor sign-in for school users (A4), separation between schools (S4, PR14), and whether a record deleted through the product's own feature stays reachable by a school account, which bears on PR13.
- Criteria answered mainly from the supplier's own records, where a test from outside sees little or nothing: where data is stored and from where staff reach it (H1, H2, H5), the continuous monitoring plan and patch timing (T1, T3), the incident response plan, its testing and breach notification (T6, T6A, T7), security testing in development (Q5), and the published sub processor list (PR17).
- Criteria that end in a school's choice, where the supplier provides the options: the authentication settings an administrator applies (A19), whether any discovery between schools is switched on where the product allows it (PR14), and which integrations a school connects, with INT7 covering the agreements behind them.
ST4S penetration testing: what criterion T1 and the evidence list ask for
Among the security criteria, penetration testing sits inside criterion T1, in the processes and testing group; the AI Module separately asks, in AI_T2, about security testing of AI features that includes penetration and jailbreak testing.
T1 asks whether the supplier runs a continuous monitoring plan across all of its systems and infrastructure, taking in automatic asset discovery, regular automated scanning for missing patches against a current vulnerability database, a penetration test after a major change or at least once a year, assessment of each vulnerability's likely impact and of fixes weighed on effectiveness, cost and the controls already in place, and fixes prioritised by risk.
It carries a hash, sits on both the Tier 1 and Tier 2 minimum criteria lists, and is asked in the Readiness Check.
Its response options distinguish an answer that meets all of these from one that also uses outside, independent testers for the penetration testing.
Section 6.8 lists the documents a supplier is required to upload, depending on its answers. Against T1 it lists EV10, a redacted copy of the service's most recent penetration test report, and EV11, redacted copies of its most recent vulnerability assessment reports.
The same section says the ST4S Team may contact an author or certifying body to check that documents are genuine, and one of its examples is asking a penetration tester for proof of their certification, such as a certificate number the Team can check.
Among its examples of why an assessment may be discontinued, section 2.2.8 lists a service that is not ready because its documentation is incomplete, it lacks a pen test, or other critical items are missing.
Criterion Q5 asks a related but different question: whether every application development for the service is assessed under a security testing method that follows the guidance in what the criterion calls the "latest" industry standard frameworks, with OWASP Testing Guide v4.2 and BSIMM given as examples.
One penetration test report speaks to the penetration testing element of T1 for one build. It does not show a monitoring plan, a scan schedule, a patching record or a development process, and it does not show that later releases were tested.
Before you commission a test for ST4S evidence
Settle these with the ST4S Team, through the contact form on its website, and with your tester before any testing is booked.
- What the ST4S Team will accept as EV10 for your product, including any certification it expects the tester to hold and whether it expects the tester to be independent of the team that builds and runs the product.
- The service and version the questionnaire names. Section 3.5 ties what a supplier may say about an outcome to the specific version of its product, so scope the test to that version and ask for the report to name the build.
- What the report records: the build, dates, roles, accounts and exclusions, with each finding's status after the retest, so the redacted copy you lodge says what was and was not tested.
- How the copy will be redacted. EV10 asks for a redacted report, so agree in advance how credentials, tokens and any personal details in the evidence will be removed.
Using the question-to-evidence worksheet as an ST4S assessment checklist
The worksheet on this page has fifteen rows.
Each one names a group of criteria by identifier and notes which of them the guide marks with a hash, puts on a minimum criteria list or asks in the Readiness Check, then gives what the criteria ask in short form, where an edtech web application meets them, the evidence to hold with a label on every item, how to produce that evidence and the row's own limits, an invented example answer, and what the evidence cannot show. The labels mean three different things.
- Supplier record: a document, register, setting or statement the supplier holds and answers from, such as a hosting configuration, an incident register or a published sub processor list.
- School decision: a choice the school, or its education authority, makes about whether and how to use the product, such as enforcing multi-factor sign-in for its staff or accepting a hosting location. The supplier offers the options; the answer is the school's.
- Independent test artefact: a dated result from testing the running product, naming the build, accounts and date, produced by a tester outside the team that builds and runs it. When your own team runs a row, file its result as a supplier record.
The order to work in
Work through the rows in this order, so nothing is sent until the staging copy and the invented schools are ready.
- Start with the read-only rows: 3, 5, 11, 12, 14 and 15. They read documents, consoles and lists and send nothing to the product.
- Before any other row, build two invented schools on a staging copy, Test School North and Test School South, with the accounts the worksheet description lists, each with an inbox and phone number that you control. No real student's, guardian's or staff member's record goes into staging, which row 5 checks.
- Run the rows that sign in (1, 2, 6, 7, 8, 9 and 10) only on that staging copy. Rows 4 and 13 send a few plain requests, with no sign-in, to the public hostnames your business owns. For every one of these rows, fix the request count before you start, counting each sign-in, send no load, and read your hosting provider's testing policy first.
- Where a row passes through a service run by another business, for example a hosted sign-in or email provider, read its terms on testing and ask the provider before sending anything to it.
- Should a response contain a real person's details, halt the run, note what appeared without copying it, and handle it under your incident response plan, which row 11 reviews.
- Write each answer from the evidence you hold. The invented examples show the shape of an answer that stays inside its evidence, and they are not wording to copy: under section 5.3 of the guide, every answer is your attestation.
Who makes each decision about an ST4S answer
The worksheet gathers evidence.
Different parties decide what it means, and the tester is not one of them.
- The ST4S Team analyses and validates the supplier's responses, may verify documents with their author, and prepares the school-level report. Under section 2.2.5 of the guide, a challenge to a finding or outcome goes first to the ST4S Team, which takes it to the Working Group where needed, and once a matter reaches the Working Group, a decision by one of its members, the Working Group or NEDAG is final.
- The education authority decides what to do with an outcome. Section 2.2.4 lets jurisdictions accept or reject it, add requirements or run their own assessment activities, and its examples include local regulations that require an education authority to carry out a privacy impact assessment, and a member that accepts only a certain risk level or only services that host data within its jurisdiction. The privacy impact assessment guide covers what such an assessment involves.
- The school decides how it uses the product: which settings it enforces, which treatments it applies, and whether it seeks parental or student consent before personal information is disclosed or shared with the service, which the guide ties to the Medium rating.
- The supplier decides what it attests to, and section 5.3 has it confirm that those answers are true and up to date.
Privacy law and other buyer requests
Privacy law is a separate question from ST4S. The APP 11 guide covers it, and whether it applies to your business is one to settle with your adviser. A school that sends its own security questionnaire instead of, or as well as, an ST4S request is covered by the security questionnaires guide, and the education industry guide follows student and family data through school workflows and tests each step.
Where Clever Ops testing fits an ST4S answer, and its limits
Clever Ops tests the parts of a product that are reached through a browser or an API.
Where teachers, students or families sign in to an edtech product, the service that matches is the web application and API penetration test, run signed in against the custom application and its API. Among other areas it covers sessions and sign-in, what every role can view or alter, whether one school's accounts can reach another school's records, direct calls to the API that bypass the screens, and file uploads.
Its scope wants two test accounts for each role and, for a platform serving several customer organisations, accounts in two of those organisations; for a schools platform that is two schools, like the invented pair in the worksheet.
A larger or multi-tenant platform is scoped on its own, after a longer call. Rows 6 to 10 of the worksheet sit on that ground.
A public marketing site with no sign-in suits the website security check, an outside review of certificates and transport security, exposed files and admin paths, headers and cookies, and any software version the site reveals; rows 4 and 13 take the same outside view. A feature built on AI, for example a writing assistant for students, is a job for the AI chatbot and agent security test; that test examines the feature and not the AI model behind it, and the Supplier Guide's AI Module carries separate criteria and a separate evidence list for such features.
Each report sets out the scope actually tested, every finding with its evidence redacted, and the checks that found nothing on that build with those accounts. One consolidated retest is included, after which each finding is given one status: fixed, or partially fixed, or still reproducible, or not retested.
Whether to lodge the report as EV10, and in what redacted form, is the supplier's decision, and whether the ST4S Team accepts it is the Team's. 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.
A Clever Ops report is evidence for the supplier to weigh. It is not an ST4S assessment, an ST4S outcome or a badge, and it cannot answer criteria that rest on the supplier's own records, such as hosting, the monitoring plan, staff screening or the incident response plan.
Clever Ops is not CREST accredited and has no other security accreditation or certification. When an education authority or a contract requires a provider with CREST accreditation, Clever Ops passes that work to one that holds it. Section 6.8 lets the ST4S Team ask a tester for proof of certification, so ask the Team what it expects of the tester before any test is booked.
If Clever Ops built the product or hosts it, the engagement is labelled a security review, since a builder or host would be checking its own work.
The report then records who built the product and who hosts it. Clever Ops leaves untested any vendor's own product or platform, a hosting, sign-in or email provider's service for example, and office networks, staff devices and red teaming need a network or red team specialist.
Frequently asked questions
No. ST4S outcomes come from the ST4S Team's assessment of a supplier's questionnaire responses and evidence, and the Supplier Guide gives them as outcomes such as Low risk, Medium risk or Use Responsibly, or as Non-compliant or another status. A penetration test report is one document a supplier may be required to upload: section 6.8 lists a redacted copy of the service's most recent one as EV10, against criterion T1. Whether a particular report is accepted is the ST4S Team's decision.
Criterion T1 asks about a continuous monitoring plan that, among other things, takes in a penetration test after a major change or at least once a year, and the AI Module asks about penetration and jailbreak testing of AI features in AI_T2. T1 carries a hash, which the Supplier Guide uses for questions where failing the minimum acceptable response results in a Non-compliant outcome or, for some items, a high risk outcome. Section 2.2.8 gives a service that is not ready because it lacks a pen test, among other gaps, as one of its examples of why an assessment may be discontinued. What will count for your product, and from which tester, is a question to put to the ST4S Team before you commission a test.
No. Section 5.7 of the Supplier Guide bars a supplier from publishing that it is in the assessment process, or from suggesting, openly or by implication, that it conforms to ST4S standards and criteria, while the only step it has finished is the Readiness Check and it holds no recent, successful full assessment of the service that is still valid. The Readiness Check page also says its results cannot be used to promote a service with current or future customers. Once a full assessment ends with a final report, section 3.5 lets a supplier tell customers it took part and meets ST4S's minimum required standards, naming the version of the standards and the version of its product assessed, while the report and the specific rating stay with the supplier.
Two years from the date of completion, according to the ST4S page on validity, with reassessment due every two years or sooner if significant changes occur that could affect compliance with current privacy, security or online safety standards. The supplier is responsible for telling ST4S about such changes. Section 2.2.6 of the Supplier Guide also gives two years, unless the assessment is withdrawn before then, and lists changes that will generally call for a reassessment, including a change of owner, a move of hosting location or infrastructure, and adding or removing AI features. Suppliers using the ST4S Product Badge also commit to a Readiness Check or a full assessment at least annually.
Readiness Check results stay private until the supplier formally submits them, when they go to the ST4S Working Group for prioritisation, according to the Readiness Check page. The general information page says final results of a full assessment go to the Working Group's members, the chief information officers of the education jurisdictions and New Zealand's Ministry of Education, and approved stakeholders may pass them on to their schools. Section 3.1 of the Supplier Guide says how and when each education authority passes findings to schools is a local matter, and section 3.2 says assessment information may also go to other parties the Working Group or NEDAG nominates. Under section 3.5, the supplier itself must not give the report, or any extract of it, to anyone outside its own organisation.
Ask whoever sent the request, because the ST4S framework versions page, read on 15 November 2026, gives two answers. Its versions list labels v2026.1 "LATEST" and v2025.1 and v2024.1 "CURRENT", while its "What's current" paragraph says the latest version is v2025.1 and versions 2024.1 and V2023.1 remain current. The same page says a version marked Latest or Active is considered by the Working Group, and a version marked Transitioning may be considered depending on the school or Working Group member.
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.