Skip to main content
Compliance guideReviewed

CPS 230: critical operations, service providers, and what a website or web application test can evidence

Prudential Standard CPS 230 Operational Risk Management asks an APRA-regulated entity to manage its operational risks, maintain its critical operations through disruptions, and manage the risks arising from service providers.

This guide reads the 2026 instrument on the Federal Register of Legislation paragraph by paragraph, sets out what it says about material service providers and their agreements, marks where APRA's guidance in CPG 230 fits, and draws the line between what a website or web application test can evidence and the operational controls that sit elsewhere. A fillable dependency map and evidence register comes with the guide.

CPS 230 critical-operations dependency map and evidence register for the web channel

Three parts in one table. Rows M1 to M8 map each critical operation the web channel supports to the public website, the customer app or portal, identity, payments, and the service providers and fourth parties behind them.

How to use it

Rows T1 to T5 are tolerance scenarios, invented for the exercise, to rehearse on paper against the tolerance levels in row M2. Rows E1 to E6 are the evidence register. Copy the table into a spreadsheet, add a line under rows M3 to M7 for each website, app and provider you have, link each one to the critical operations in row M1, and put real dates and owners in every row. Only row M8 sends requests. Its limits are printed on the row itself, starting with this one: a site and app your business runs or holds written authority to test, on a staging copy where one exists. Rows T1 to T5 are talked through on paper, and the remaining rows read or write records and send nothing. A system a vendor runs is tested or changed only within that vendor's own terms, so read them first and take any gap to the vendor. In each row, the paragraph column quotes Prudential Standard CPS 230 as made by determination No. 1 of 2026, or shortens it closely, with the paragraph number. CPG 230 material in any row is guidance, not a requirement. Two columns mark the boundary: the evidence a web test can offer, and what sits elsewhere. The register records evidence. Whether any of it meets CPS 230 is weighed by the APRA-regulated entity, its Board, its internal audit function and APRA; for a supplier, by its regulated customer.

19 rows

CPS 230 critical-operations dependency map and evidence register for the web channel

Row and recordCPS 230 text, by paragraph numberWhat to map or recordEvidence a web test can offerWhat sits elsewhere, and who decidesLimits of this row
M1. Critical operations the web channel supportsParagraph 34: critical operations are processes undertaken by an APRA-regulated entity or its service provider which, if disrupted beyond tolerance levels, would have a material adverse impact on its depositors, policyholders, beneficiaries or other customers, or its role in the financial system. Paragraph 33(a): define, identify and maintain a register of its critical operations. Paragraph 35 sets the operations an entity must, at a minimum, classify as critical unless it can justify otherwise, including, for all APRA-regulated entities, customer enquiries and the systems and infrastructure needed to support critical operations.Each critical operation in the entity's register that a website, customer app or API delivers or supports, under the register's own name, with the person who owns the operation and the web components it uses (rows M3 to M6). A supplier fills this row from what its customer tells it: which of the customer's critical operations its service supports, if any.Nothing towards the classification. A test report's scope as tested names the application, routes and roles tested, so the map can show which critical operation a test touched.Which operations are critical is the entity's classification under paragraphs 33 to 35. Paragraph 36 says APRA may require an APRA-regulated entity, or a class of APRA-regulated entities, to classify a business operation as a critical operation.Read-only row: it copies from the entity's register and sends no requests.
M2. Tolerance levels for each of those operationsParagraph 37: for each critical operation, establish tolerance levels for (a) the maximum period of time the entity would tolerate a disruption to the operation; (b) the maximum extent of data loss the entity would accept as a result of a disruption; and (c) minimum service levels the entity would maintain while operating under alternative arrangements during a disruption. Paragraph 21(b): the Board must approve the BCP and tolerance levels for disruptions to critical operations.The three levels as the entity set them, the date the Board approved them, and, for each web component in the map, whether its recovery arrangements were planned against them. CPG 230, as guidance, lists recovery time objectives among the factors to consider in setting a maximum period, a recovery point objective for maximum data loss, and a recovery level objective for minimum service levels.Nothing. A security test does not measure recovery time, data loss or service levels.Setting the levels is the entity's and approving them the Board's. Paragraph 38 says APRA may require an entity to review and change its tolerance levels for a critical operation, and may set tolerance levels where it identifies a heightened risk or material weakness.Read-only row: it copies from the BCP and sends no requests.
M3. Public website and enquiry formsParagraph 26(b): identify and document the processes and resources needed to deliver critical operations, including people, technology, information, facilities and service providers, the interdependencies across them, and the associated risks, obligations, key data and controls. Paragraph 35(d) names customer enquiries among the operations every APRA-regulated entity must, at a minimum, classify as critical unless it can justify otherwise.Every hostname the public site and its enquiry forms use; the host, domain registrar, DNS provider and any content delivery network; where each form's submissions go, such as an inbox, a customer relationship tool or a case system; who can change the site and its DNS; and which critical operation, if any, each form feeds.A Clever Ops website security check reports a hostname inventory, including names you may not know about, with where each one points and who appears to run it, and checks certificates and records that point at third-party services no longer in use. Those results describe the hostnames checked, on the dates checked.Whether the site can be kept or restored within tolerance is a business continuity question for the BCP and its testing under paragraphs 33 and 42. Load and denial-of-service testing are not part of any Clever Ops service.Read-only row: it lists entries from your DNS, hosting and form settings, opens no screen that shows a password, key or secret, and sends no requests to the site.
M4. Customer app or portal and its APIParagraph 26(b), as in M3. Paragraph 24: maintain appropriate and sound information and IT capability to meet current and projected business requirements and to support critical operations and risk management; in managing technology risks, monitor the age and health of its information assets and meet the requirements for information security in CPS 234.The app's parts (front end, API, database, file storage, background jobs and admin screens), who runs each one, the framework and major components with the date each was last updated, the kinds of data each part holds, and the critical operations each part supports.A Clever Ops web application and API penetration test reports, for the build and accounts tested, findings on sign-in, sessions, access between customers and between roles, the API called directly, business rules and file uploads, with negative results for checks that found nothing. Older API versions that still answer are reported.The age and health of information assets across the whole estate, and whether IT capability is sound, are the entity's assessments under paragraph 24. The information security requirements paragraph 24 points to are in CPS 234, which has its own guide in this library.Read-only row: it records architecture and ownership from your own documents and sends no requests.
M5. Identity: sign-in, account recovery and administrator accessParagraph 26(b), as in M3. Paragraph 23: manage the full range of operational risks, including but not limited to legal risk, regulatory risk, compliance risk, conduct risk, technology risk, data risk and change management risk.Which service signs customers and staff in, whether built into the app or a separate identity provider; how accounts are recovered; which service sends one-time codes; every administrator account for the app, hosting, DNS and identity service; and what customers can still do while sign-in is unavailable.A web application test checks sign-up, sign-in, password reset and session ending on the application's side, from test accounts, and whether a lower role can reach administrator functions. Lockout and reset checks run only on test accounts, where the rules of engagement allow them.Clever Ops does not test the identity provider's own platform, and its availability is the provider's; this row reviews only your configuration in it, as far as its terms allow. Staff devices and office networks go to a network specialist.Read-only row: it reads settings and lists accounts, opens no screen that shows a password, key or secret, and sends no requests.
M6. PaymentsParagraph 35(a): for an ADI, payments are among the business operations the entity must, at a minimum, classify as critical operations, unless it can justify otherwise. Paragraph 26(b), as in M3.The payment provider, the hosted checkout or payment page, the route in your app that receives the provider's notifications, how a payment is marked complete, and what happens to an order or instruction when the provider cannot be reached.A web application test covers your application's side of a payment: whether the amount and plan come from the server, whether an upgrade is applied only once the provider confirms the payment, and whether the notification route rejects a message without a valid signature. Payment checks run on a staging copy linked to a non-live payment account wherever one is available.Clever Ops does not test the payment provider's own platform. PCI DSS scans and assessments are referred to providers that hold the relevant role, because Clever Ops is not a PCI Approved Scanning Vendor or a Qualified Security Assessor.Read-only row: it records the payment flow from your own documents and settings and sends no requests.
M7. Service providers and fourth parties behind each componentParagraph 48: identify and maintain a register of material service providers and manage the material risks associated with using them; material service providers are those on which the entity relies to undertake a critical operation or that expose it to material operational risk. Paragraph 49: an APRA-regulated entity must, at a minimum, classify a provider of the services it lists as a material service provider, unless it can justify otherwise, and for all APRA-regulated entities those listed services are risk management, core technology services and internal audit (49(d)). Footnote 14 to paragraph 47(c): a fourth party is a party that a service provider relies on in delivering services to an APRA-regulated entity.For each component in rows M3 to M6: who provides it, the critical operation it supports, whether the entity classifies the provider as material and who made that decision, and the fourth parties the provider relies on, such as its own host or email, identity or payment services. CPG 230, as guidance, says that where a material arrangement is relied on to deliver a critical operation, a prudent entity would take reasonable steps to list fourth parties involved in delivery of the critical operation.The hostname inventory from a website security check shows which outside services answer on your domain, and row M8 lists the outside hosts a page loads. Neither shows whether a provider is material.Materiality is the entity's classification under paragraphs 48 and 49. Paragraph 51 says APRA may require an entity, or a class of entities, to classify a service provider, type of service provider or service provider arrangement as material.Read-only row: it compares your supplier list with rows M3 to M6 and sends no requests.
M8. What the pages actually loadParagraph 26(b): the interdependencies across the people, technology, information, facilities and service providers needed to deliver critical operations. Paragraph 47(c): the service provider management policy must include the entity's approach to managing the risks associated with any fourth parties that material service providers rely on to deliver a critical operation.Open the public home page, one enquiry form page and the app's sign-in page once each in a browser with its developer tools open on the network view. Read only the host name of each request the page made, without opening any request. Compare that list with rows M3 to M7, and add any host nobody had listed, with who runs it and why the page calls it.Nothing from a tester: this row is a quick check your own team can run between tests. A website security check also covers what the site hands to every browser, including keys and secrets in page source and JavaScript.The list shows what loaded on those pages on that day. It says nothing about what each outside service does with data, how resilient it is, or what loads behind sign-in beyond the sign-in page.Sends requests. Point it only at a site and app that your business runs, or that it holds written authority to test, and use a staging copy where one exists. Sign in to nothing and submit no form. No load: three page loads in total, one at a time, with no reloads. Each load also fetches the files that page itself calls for, such as scripts, styles, images and fonts, and those fetches are what the row reads; send nothing else. Note the host name of each request and nothing more, and do not open a request's details, headers, cookies or body. Keep passwords, keys, tokens, cookie values, request bodies and any link that works for whoever holds it, such as an invitation, reset or file link, out of the map and out of any screenshot. If one of the pages is run by a vendor, such as a hosted sign-in page, check that vendor's terms before loading it, raise anything the terms leave open with the vendor, and load the page only within what the terms or the vendor's reply permit. If a page displays someone else's real data, halt: note where and when, open nothing else, and tell the person in your business who handles privacy and incidents. Requests in this row: three page loads in total, one of each page listed, each with the files that page calls for and nothing more. The row signs in to no account, so it makes no test-account sign-ins.
T1. Scenario: the public website or enquiry channel is unavailableParagraph 26(c): undertake scenario analysis to identify and assess the potential impact of severe operational risk events, test operational resilience and identify the need for new or amended controls and other mitigation strategies. Paragraph 42: a systematic testing program for the BCP that covers all critical operations and includes an annual business continuity exercise, testing the BCP's effectiveness and its ability to meet tolerance levels in a range of severe but plausible scenarios.The scenario: the hosting account, DNS or certificate behind the public site fails, and customers cannot lodge enquiries online. Record how enquiries reach the business instead, how long the alternative takes to switch on, which enquiries could be lost, and whether each result sits within the row M2 tolerances for the operation the site supports.A website security check can show weaknesses on the tested hostnames, such as certificate problems or records pointing at services no longer in use, that the scenario can take as its starting point. It does not show how long recovery takes.Whether the alternative meets tolerance is shown by the BCP exercise, not by a security test. Under paragraph 21(b), the Board reviews the results of testing and oversees the execution of any findings.Paper exercise: it is rehearsed around a table, switches nothing off, changes no live system and sends no requests. The scenario is invented for the exercise.
T2. Scenario: a security finding forces part of the customer app offlineParagraph 13: to the extent practicable, prevent disruption to critical operations, adapt processes and systems to continue to operate within tolerance levels in the event of a disruption and return to normal operations promptly once a disruption is over. Paragraph 41: notify APRA as soon as possible, and not later than 24 hours after, if it has suffered a disruption to a critical operation outside tolerance.The invented scenario: a finding shows that one customer can open another customer's records, and the business takes that feature offline until a fix ships. Record which critical operation would lose the feature, how customers would be served meanwhile, how long the fix and its retest would take against the maximum-period tolerance, and who would make the call to take the feature offline.For a real finding, a test report gives the facts the scenario needs: the affected routes, the build, the business-risk rating and the retest criteria the fix has to pass.Whether a disruption is outside tolerance is measured against the tolerance levels the entity set under paragraph 37, and that measurement is the entity's. Footnote 11 says a notification of an information security incident reported under CPS 234 does not need to be separately reported under the notification requirements of CPS 230.Paper exercise: it is rehearsed around a table, switches nothing off, changes no live system and sends no requests. The scenario is invented for the exercise.
T3. Scenario: customers and staff cannot sign inParagraph 37(c): minimum service levels the entity would maintain while operating under alternative arrangements during a disruption. Paragraph 39(c): the BCP must include the actions the entity would take to maintain its critical operations within tolerance levels through disruptions.The scenario: the sign-in service is down for a day. Record what customers can still do without signing in, how staff reach records, whether a fallback sign-in route exists and who can switch it on, and how that route is closed again afterwards.If a fallback sign-in route is planned, a web application test can check it like any other sign-in flow, from test accounts, before it is ever needed.The identity provider's resilience is the provider's, and Clever Ops does not test its platform. The entity's arrangement with it falls under its service provider management policy, which paragraph 46 says the entity must maintain.Paper exercise: it is rehearsed around a table, switches nothing off, changes no live system and sends no requests. The scenario is invented for the exercise.
T4. Scenario: payment notifications stop arrivingParagraph 37(b): the maximum extent of data loss the entity would accept as a result of a disruption. Paragraph 43: the testing program must include a range of severe but plausible scenarios, including disruptions to services provided by material service providers and scenarios where contingency arrangements are required.The scenario: the payment provider's notifications stop for several hours, then resume. Record which orders, instructions or payments would be left in an unknown state, how staff would reconcile them, and whether the data that could be lost stays within the maximum data loss tolerance.A web application test checks, on a staging copy where one is available, whether a payment is marked complete only once the provider confirms it, and whether the notification route rejects a message without a valid signature. The result tells the exercise how the tested build marks payments while confirmations are missing.Disruption at the provider, and reconciliation afterwards, are business continuity matters. CPG 230, as guidance, says entities that rely on material service providers would seek to confirm that those providers also maintain robust BCP testing, and that joint testing of arrangements with the service provider could be considered.Paper exercise: it is rehearsed around a table, switches nothing off, changes no live system and sends no requests. The scenario is invented for the exercise.
T5. Scenario: the host or core technology provider fails or is exitedParagraph 55: for each material arrangement, (a) identify and manage risks that could affect the ability of the service provider to provide the service on an ongoing basis; (b) identify and manage risks to the entity that could result from the arrangement, such as step-in risk or contagion risk; (c) ensure it can execute its BCP if needed; and (d) ensure it can conduct an orderly exit from the arrangement if needed.The scenario: the host or core technology provider behind the app stops service, or the entity decides to leave it. Record where current copies of the app's code, data and configuration are held outside that provider, the date a restore from them was last tried, and how long a move to another provider would take against the row M2 tolerances.Nothing. A security test does not test backups, restores or an exit.Exit and continuity planning fall to the entity under paragraph 55. For the provider's part in an exit, read the agreement between them and take questions about it to each party's own adviser. Under paragraph 57, paragraphs 53, 54, 55(d), 59(a) and 59(c) need not be complied with for a material arrangement where the provider falls within a category in the Attachment and the arrangement uses standardised terms or is not documented in a formal agreement.Paper exercise: it is rehearsed around a table, switches nothing off, changes no live system and sends no requests. The scenario is invented for the exercise.
E1. Control testing for the web channelParagraph 29: regularly monitor, review and test controls for design and operating effectiveness, the frequency of which must be commensurate with the materiality of the risks being controlled; the results of testing must be reported to senior management and any gaps or deficiencies in the control environment must be rectified in a timely manner.For each web component: the controls tested, by whom, when, at what scope and depth, where the report is kept, the date the results went to senior management, and the next test date with the reason for its timing. CPG 230, as guidance, says testing would be conducted by staff and teams independent of those with operational responsibility for controls.A dated report about one application's controls on the build tested: the scope as tested, findings, a CVSS v3.1 score and a business-risk rating for each rated finding, negative results, and test cases that could not run.Whether the frequency is commensurate with the materiality of the risk, and whether a gap was rectified in a timely manner, are the entity's judgements. Controls outside the test's scope, such as backups, networks and staff devices, need their own testing.Read-only row: it logs test plans, report locations and dates, and sends no requests.
E2. Remediation and the operational risk profileParagraph 30: remediate material weaknesses in operational risk management, including control gaps, weaknesses and failures; the remediation must be supported by clear accountabilities and assurance and address the root causes of weaknesses in a timely manner; identified control gaps, weaknesses and failures are included in the operational risk profile until remediated.Each finding not marked fixed: its owner, its target date, the root cause your team identified, and the dates it was entered in the operational risk profile and, later, removed from it.After the one consolidated retest, a Clever Ops report marks each finding fixed, partially fixed, still reproducible or not retested, on an identified build.Whether a finding is a material weakness, and when it counts as remediated in the risk profile, are the entity's decisions. A finding marked not retested has not been confirmed fixed.Read-only row: it tracks owners, target dates and retest statuses, and sends no requests.
E3. Incidents, near misses and the notification clocksParagraph 31: ensure that operational risk incidents and near misses are identified, escalated, recorded and addressed in a timely manner. Paragraph 32: notify APRA as soon as possible, and not later than 72 hours, after becoming aware of an operational risk incident that the entity determines to be likely to have a material financial impact or a material impact on its ability to maintain its critical operations. Paragraph 41: notify APRA as soon as possible, and not later than 24 hours after, if it has suffered a disruption to a critical operation outside tolerance.Who decides whether an event involving the web channel is an operational risk incident or a near miss, who notifies APRA and which clock applies, and the date each decision was made.Nothing. A test finding is evidence about a weakness on a tested build. It is not an incident report or a notification.Each determination is the entity's. Footnote 11 says a notification of an information security incident reported under CPS 234 does not need to be separately reported under the notification requirements of CPS 230. Where personal information is involved, the Notifiable Data Breaches scheme has its own guide in this library.Read-only row: it logs who decided what, and when, and sends no requests.
E4. The agreement with each material service providerParagraph 53: for all material arrangements, maintain a formal legally binding agreement that, at a minimum, (a) specifies the services and associated service levels; (b) sets out each party's rights, responsibilities and expectations, including in relation to the ownership of assets, ownership and control of data, dispute resolution, audit access, liability and indemnity; (c) includes provisions to ensure the entity can meet its legal and compliance obligations; (d) requires notification by the provider of its use of other material service providers it materially relies upon in providing the service to the entity through sub-contracting or other arrangements; (e) makes the provider responsible for liability for any sub-contractor's failure; (f) includes a force majeure provision indicating which parts of the contract would continue in the case of a force majeure event; and (g) includes termination provisions including, but not limited to, the right to terminate the arrangement in its entirety or parts of it, with a further termination requirement for RSE licensees. Paragraph 54: provisions allowing APRA access to documentation, data and any other information related to the service, an on-site visit to the provider, and the provider's agreement not to impede APRA in fulfilling its duties as prudential regulator.For each material arrangement in row M7: where the agreement is filed, the clause that covers each item in paragraphs 53 and 54 or a note that it is missing, whether paragraph 57 is relied on and why, and the next review date.Nothing towards the clauses themselves. Where an agreement asks the provider for security testing evidence, a report and its retest status are material the provider could share, as far as the agreement allows.Whether an agreement meets paragraphs 53 and 54 is for the entity and its own advisers. A supplier reads its agreement with the customer and asks its own adviser what the agreement requires of it.Read-only row: it works from the signed agreements and sends no requests.
E5. Monitoring each material arrangementParagraph 59: monitor and ensure that senior management receive reporting on material arrangements commensurate with the nature and usage of the service, including a regular assessment of (a) performance under the service agreement with reference to agreed service levels; (b) the effectiveness of controls to manage the risks associated with the use of the service provider; and (c) compliance of both parties with the service provider agreement.For each material arrangement: the date of each regular assessment, the service levels met or missed, the evidence of control effectiveness received, the issues raised, and when they were reported to senior management. CPG 230, as guidance, says periodic reviews could include assessment of operational issues (including information security incidents and service disruptions), control effectiveness, information security capabilities and business continuity capabilities, strategic changes and comparisons to other offerings in the market.A provider's test report and retest status can be part of the evidence of control effectiveness, for the systems the report covers.Whether the evidence shows effective controls for paragraph 59(b) is the entity's assessment.Read-only row: it logs each assessment and the reporting that followed, and sends no requests.
E6. Supplier: answering an APRA-regulated customerThe paragraphs in this map are addressed to the APRA-regulated entity. Paragraph 14: an APRA-regulated entity must not rely on a service provider unless it can ensure that in doing so it can continue to meet its prudential obligations in full and effectively manage the associated risks. Paragraph 52: before entering into or materially modifying a material arrangement, (a) undertake appropriate due diligence, including an appropriate selection process and an assessment of the provider's ability to provide the service on an ongoing basis; and (b) assess the financial and non-financial risks from reliance on the provider, including risks associated with geographic location or concentration of the provider or parties it relies on in providing the service.The customer's exact request, whether a questionnaire, a due diligence request, a contract clause or an invitation into a BCP exercise; whether the customer says you are a material service provider; the evidence you sent for each item; how much of it the agreement allows you to pass on; and the date sent.Evidence with a date and a stated scope, covering the systems you operate for that customer, and a retest status against each finding. A Clever Ops report on a system Clever Ops built or hosts also says so.Your customer judges whether the evidence is enough. As for what your agreement with that customer asks of you, read it and put the question to your own adviser.Read-only row: it pairs each item in a customer's request with a document and sends no requests.

Row and record

CPS 230 text, by paragraph number

  1. M1. Critical operations the web channel supports
    CPS 230 text, by paragraph number
    Paragraph 34: critical operations are processes undertaken by an APRA-regulated entity or its service provider which, if disrupted beyond tolerance levels, would have a material adverse impact on its depositors, policyholders, beneficiaries or other customers, or its role in the financial system. Paragraph 33(a): define, identify and maintain a register of its critical operations. Paragraph 35 sets the operations an entity must, at a minimum, classify as critical unless it can justify otherwise, including, for all APRA-regulated entities, customer enquiries and the systems and infrastructure needed to support critical operations.
    What to map or record
    Each critical operation in the entity's register that a website, customer app or API delivers or supports, under the register's own name, with the person who owns the operation and the web components it uses (rows M3 to M6). A supplier fills this row from what its customer tells it: which of the customer's critical operations its service supports, if any.
    Evidence a web test can offer
    Nothing towards the classification. A test report's scope as tested names the application, routes and roles tested, so the map can show which critical operation a test touched.
    What sits elsewhere, and who decides
    Which operations are critical is the entity's classification under paragraphs 33 to 35. Paragraph 36 says APRA may require an APRA-regulated entity, or a class of APRA-regulated entities, to classify a business operation as a critical operation.
    Limits of this row
    Read-only row: it copies from the entity's register and sends no requests.
  2. M2. Tolerance levels for each of those operations
    CPS 230 text, by paragraph number
    Paragraph 37: for each critical operation, establish tolerance levels for (a) the maximum period of time the entity would tolerate a disruption to the operation; (b) the maximum extent of data loss the entity would accept as a result of a disruption; and (c) minimum service levels the entity would maintain while operating under alternative arrangements during a disruption. Paragraph 21(b): the Board must approve the BCP and tolerance levels for disruptions to critical operations.
    What to map or record
    The three levels as the entity set them, the date the Board approved them, and, for each web component in the map, whether its recovery arrangements were planned against them. CPG 230, as guidance, lists recovery time objectives among the factors to consider in setting a maximum period, a recovery point objective for maximum data loss, and a recovery level objective for minimum service levels.
    Evidence a web test can offer
    Nothing. A security test does not measure recovery time, data loss or service levels.
    What sits elsewhere, and who decides
    Setting the levels is the entity's and approving them the Board's. Paragraph 38 says APRA may require an entity to review and change its tolerance levels for a critical operation, and may set tolerance levels where it identifies a heightened risk or material weakness.
    Limits of this row
    Read-only row: it copies from the BCP and sends no requests.
  3. M3. Public website and enquiry forms
    CPS 230 text, by paragraph number
    Paragraph 26(b): identify and document the processes and resources needed to deliver critical operations, including people, technology, information, facilities and service providers, the interdependencies across them, and the associated risks, obligations, key data and controls. Paragraph 35(d) names customer enquiries among the operations every APRA-regulated entity must, at a minimum, classify as critical unless it can justify otherwise.
    What to map or record
    Every hostname the public site and its enquiry forms use; the host, domain registrar, DNS provider and any content delivery network; where each form's submissions go, such as an inbox, a customer relationship tool or a case system; who can change the site and its DNS; and which critical operation, if any, each form feeds.
    Evidence a web test can offer
    A Clever Ops website security check reports a hostname inventory, including names you may not know about, with where each one points and who appears to run it, and checks certificates and records that point at third-party services no longer in use. Those results describe the hostnames checked, on the dates checked.
    What sits elsewhere, and who decides
    Whether the site can be kept or restored within tolerance is a business continuity question for the BCP and its testing under paragraphs 33 and 42. Load and denial-of-service testing are not part of any Clever Ops service.
    Limits of this row
    Read-only row: it lists entries from your DNS, hosting and form settings, opens no screen that shows a password, key or secret, and sends no requests to the site.
  4. M4. Customer app or portal and its API
    CPS 230 text, by paragraph number
    Paragraph 26(b), as in M3. Paragraph 24: maintain appropriate and sound information and IT capability to meet current and projected business requirements and to support critical operations and risk management; in managing technology risks, monitor the age and health of its information assets and meet the requirements for information security in CPS 234.
    What to map or record
    The app's parts (front end, API, database, file storage, background jobs and admin screens), who runs each one, the framework and major components with the date each was last updated, the kinds of data each part holds, and the critical operations each part supports.
    Evidence a web test can offer
    A Clever Ops web application and API penetration test reports, for the build and accounts tested, findings on sign-in, sessions, access between customers and between roles, the API called directly, business rules and file uploads, with negative results for checks that found nothing. Older API versions that still answer are reported.
    What sits elsewhere, and who decides
    The age and health of information assets across the whole estate, and whether IT capability is sound, are the entity's assessments under paragraph 24. The information security requirements paragraph 24 points to are in CPS 234, which has its own guide in this library.
    Limits of this row
    Read-only row: it records architecture and ownership from your own documents and sends no requests.
  5. M5. Identity: sign-in, account recovery and administrator access
    CPS 230 text, by paragraph number
    Paragraph 26(b), as in M3. Paragraph 23: manage the full range of operational risks, including but not limited to legal risk, regulatory risk, compliance risk, conduct risk, technology risk, data risk and change management risk.
    What to map or record
    Which service signs customers and staff in, whether built into the app or a separate identity provider; how accounts are recovered; which service sends one-time codes; every administrator account for the app, hosting, DNS and identity service; and what customers can still do while sign-in is unavailable.
    Evidence a web test can offer
    A web application test checks sign-up, sign-in, password reset and session ending on the application's side, from test accounts, and whether a lower role can reach administrator functions. Lockout and reset checks run only on test accounts, where the rules of engagement allow them.
    What sits elsewhere, and who decides
    Clever Ops does not test the identity provider's own platform, and its availability is the provider's; this row reviews only your configuration in it, as far as its terms allow. Staff devices and office networks go to a network specialist.
    Limits of this row
    Read-only row: it reads settings and lists accounts, opens no screen that shows a password, key or secret, and sends no requests.
  6. M6. Payments
    CPS 230 text, by paragraph number
    Paragraph 35(a): for an ADI, payments are among the business operations the entity must, at a minimum, classify as critical operations, unless it can justify otherwise. Paragraph 26(b), as in M3.
    What to map or record
    The payment provider, the hosted checkout or payment page, the route in your app that receives the provider's notifications, how a payment is marked complete, and what happens to an order or instruction when the provider cannot be reached.
    Evidence a web test can offer
    A web application test covers your application's side of a payment: whether the amount and plan come from the server, whether an upgrade is applied only once the provider confirms the payment, and whether the notification route rejects a message without a valid signature. Payment checks run on a staging copy linked to a non-live payment account wherever one is available.
    What sits elsewhere, and who decides
    Clever Ops does not test the payment provider's own platform. PCI DSS scans and assessments are referred to providers that hold the relevant role, because Clever Ops is not a PCI Approved Scanning Vendor or a Qualified Security Assessor.
    Limits of this row
    Read-only row: it records the payment flow from your own documents and settings and sends no requests.
  7. M7. Service providers and fourth parties behind each component
    CPS 230 text, by paragraph number
    Paragraph 48: identify and maintain a register of material service providers and manage the material risks associated with using them; material service providers are those on which the entity relies to undertake a critical operation or that expose it to material operational risk. Paragraph 49: an APRA-regulated entity must, at a minimum, classify a provider of the services it lists as a material service provider, unless it can justify otherwise, and for all APRA-regulated entities those listed services are risk management, core technology services and internal audit (49(d)). Footnote 14 to paragraph 47(c): a fourth party is a party that a service provider relies on in delivering services to an APRA-regulated entity.
    What to map or record
    For each component in rows M3 to M6: who provides it, the critical operation it supports, whether the entity classifies the provider as material and who made that decision, and the fourth parties the provider relies on, such as its own host or email, identity or payment services. CPG 230, as guidance, says that where a material arrangement is relied on to deliver a critical operation, a prudent entity would take reasonable steps to list fourth parties involved in delivery of the critical operation.
    Evidence a web test can offer
    The hostname inventory from a website security check shows which outside services answer on your domain, and row M8 lists the outside hosts a page loads. Neither shows whether a provider is material.
    What sits elsewhere, and who decides
    Materiality is the entity's classification under paragraphs 48 and 49. Paragraph 51 says APRA may require an entity, or a class of entities, to classify a service provider, type of service provider or service provider arrangement as material.
    Limits of this row
    Read-only row: it compares your supplier list with rows M3 to M6 and sends no requests.
  8. M8. What the pages actually load
    CPS 230 text, by paragraph number
    Paragraph 26(b): the interdependencies across the people, technology, information, facilities and service providers needed to deliver critical operations. Paragraph 47(c): the service provider management policy must include the entity's approach to managing the risks associated with any fourth parties that material service providers rely on to deliver a critical operation.
    What to map or record
    Open the public home page, one enquiry form page and the app's sign-in page once each in a browser with its developer tools open on the network view. Read only the host name of each request the page made, without opening any request. Compare that list with rows M3 to M7, and add any host nobody had listed, with who runs it and why the page calls it.
    Evidence a web test can offer
    Nothing from a tester: this row is a quick check your own team can run between tests. A website security check also covers what the site hands to every browser, including keys and secrets in page source and JavaScript.
    What sits elsewhere, and who decides
    The list shows what loaded on those pages on that day. It says nothing about what each outside service does with data, how resilient it is, or what loads behind sign-in beyond the sign-in page.
    Limits of this row
    Sends requests. Point it only at a site and app that your business runs, or that it holds written authority to test, and use a staging copy where one exists. Sign in to nothing and submit no form. No load: three page loads in total, one at a time, with no reloads. Each load also fetches the files that page itself calls for, such as scripts, styles, images and fonts, and those fetches are what the row reads; send nothing else. Note the host name of each request and nothing more, and do not open a request's details, headers, cookies or body. Keep passwords, keys, tokens, cookie values, request bodies and any link that works for whoever holds it, such as an invitation, reset or file link, out of the map and out of any screenshot. If one of the pages is run by a vendor, such as a hosted sign-in page, check that vendor's terms before loading it, raise anything the terms leave open with the vendor, and load the page only within what the terms or the vendor's reply permit. If a page displays someone else's real data, halt: note where and when, open nothing else, and tell the person in your business who handles privacy and incidents. Requests in this row: three page loads in total, one of each page listed, each with the files that page calls for and nothing more. The row signs in to no account, so it makes no test-account sign-ins.
  9. T1. Scenario: the public website or enquiry channel is unavailable
    CPS 230 text, by paragraph number
    Paragraph 26(c): undertake scenario analysis to identify and assess the potential impact of severe operational risk events, test operational resilience and identify the need for new or amended controls and other mitigation strategies. Paragraph 42: a systematic testing program for the BCP that covers all critical operations and includes an annual business continuity exercise, testing the BCP's effectiveness and its ability to meet tolerance levels in a range of severe but plausible scenarios.
    What to map or record
    The scenario: the hosting account, DNS or certificate behind the public site fails, and customers cannot lodge enquiries online. Record how enquiries reach the business instead, how long the alternative takes to switch on, which enquiries could be lost, and whether each result sits within the row M2 tolerances for the operation the site supports.
    Evidence a web test can offer
    A website security check can show weaknesses on the tested hostnames, such as certificate problems or records pointing at services no longer in use, that the scenario can take as its starting point. It does not show how long recovery takes.
    What sits elsewhere, and who decides
    Whether the alternative meets tolerance is shown by the BCP exercise, not by a security test. Under paragraph 21(b), the Board reviews the results of testing and oversees the execution of any findings.
    Limits of this row
    Paper exercise: it is rehearsed around a table, switches nothing off, changes no live system and sends no requests. The scenario is invented for the exercise.
  10. T2. Scenario: a security finding forces part of the customer app offline
    CPS 230 text, by paragraph number
    Paragraph 13: to the extent practicable, prevent disruption to critical operations, adapt processes and systems to continue to operate within tolerance levels in the event of a disruption and return to normal operations promptly once a disruption is over. Paragraph 41: notify APRA as soon as possible, and not later than 24 hours after, if it has suffered a disruption to a critical operation outside tolerance.
    What to map or record
    The invented scenario: a finding shows that one customer can open another customer's records, and the business takes that feature offline until a fix ships. Record which critical operation would lose the feature, how customers would be served meanwhile, how long the fix and its retest would take against the maximum-period tolerance, and who would make the call to take the feature offline.
    Evidence a web test can offer
    For a real finding, a test report gives the facts the scenario needs: the affected routes, the build, the business-risk rating and the retest criteria the fix has to pass.
    What sits elsewhere, and who decides
    Whether a disruption is outside tolerance is measured against the tolerance levels the entity set under paragraph 37, and that measurement is the entity's. Footnote 11 says a notification of an information security incident reported under CPS 234 does not need to be separately reported under the notification requirements of CPS 230.
    Limits of this row
    Paper exercise: it is rehearsed around a table, switches nothing off, changes no live system and sends no requests. The scenario is invented for the exercise.
  11. T3. Scenario: customers and staff cannot sign in
    CPS 230 text, by paragraph number
    Paragraph 37(c): minimum service levels the entity would maintain while operating under alternative arrangements during a disruption. Paragraph 39(c): the BCP must include the actions the entity would take to maintain its critical operations within tolerance levels through disruptions.
    What to map or record
    The scenario: the sign-in service is down for a day. Record what customers can still do without signing in, how staff reach records, whether a fallback sign-in route exists and who can switch it on, and how that route is closed again afterwards.
    Evidence a web test can offer
    If a fallback sign-in route is planned, a web application test can check it like any other sign-in flow, from test accounts, before it is ever needed.
    What sits elsewhere, and who decides
    The identity provider's resilience is the provider's, and Clever Ops does not test its platform. The entity's arrangement with it falls under its service provider management policy, which paragraph 46 says the entity must maintain.
    Limits of this row
    Paper exercise: it is rehearsed around a table, switches nothing off, changes no live system and sends no requests. The scenario is invented for the exercise.
  12. T4. Scenario: payment notifications stop arriving
    CPS 230 text, by paragraph number
    Paragraph 37(b): the maximum extent of data loss the entity would accept as a result of a disruption. Paragraph 43: the testing program must include a range of severe but plausible scenarios, including disruptions to services provided by material service providers and scenarios where contingency arrangements are required.
    What to map or record
    The scenario: the payment provider's notifications stop for several hours, then resume. Record which orders, instructions or payments would be left in an unknown state, how staff would reconcile them, and whether the data that could be lost stays within the maximum data loss tolerance.
    Evidence a web test can offer
    A web application test checks, on a staging copy where one is available, whether a payment is marked complete only once the provider confirms it, and whether the notification route rejects a message without a valid signature. The result tells the exercise how the tested build marks payments while confirmations are missing.
    What sits elsewhere, and who decides
    Disruption at the provider, and reconciliation afterwards, are business continuity matters. CPG 230, as guidance, says entities that rely on material service providers would seek to confirm that those providers also maintain robust BCP testing, and that joint testing of arrangements with the service provider could be considered.
    Limits of this row
    Paper exercise: it is rehearsed around a table, switches nothing off, changes no live system and sends no requests. The scenario is invented for the exercise.
  13. T5. Scenario: the host or core technology provider fails or is exited
    CPS 230 text, by paragraph number
    Paragraph 55: for each material arrangement, (a) identify and manage risks that could affect the ability of the service provider to provide the service on an ongoing basis; (b) identify and manage risks to the entity that could result from the arrangement, such as step-in risk or contagion risk; (c) ensure it can execute its BCP if needed; and (d) ensure it can conduct an orderly exit from the arrangement if needed.
    What to map or record
    The scenario: the host or core technology provider behind the app stops service, or the entity decides to leave it. Record where current copies of the app's code, data and configuration are held outside that provider, the date a restore from them was last tried, and how long a move to another provider would take against the row M2 tolerances.
    Evidence a web test can offer
    Nothing. A security test does not test backups, restores or an exit.
    What sits elsewhere, and who decides
    Exit and continuity planning fall to the entity under paragraph 55. For the provider's part in an exit, read the agreement between them and take questions about it to each party's own adviser. Under paragraph 57, paragraphs 53, 54, 55(d), 59(a) and 59(c) need not be complied with for a material arrangement where the provider falls within a category in the Attachment and the arrangement uses standardised terms or is not documented in a formal agreement.
    Limits of this row
    Paper exercise: it is rehearsed around a table, switches nothing off, changes no live system and sends no requests. The scenario is invented for the exercise.
  14. E1. Control testing for the web channel
    CPS 230 text, by paragraph number
    Paragraph 29: regularly monitor, review and test controls for design and operating effectiveness, the frequency of which must be commensurate with the materiality of the risks being controlled; the results of testing must be reported to senior management and any gaps or deficiencies in the control environment must be rectified in a timely manner.
    What to map or record
    For each web component: the controls tested, by whom, when, at what scope and depth, where the report is kept, the date the results went to senior management, and the next test date with the reason for its timing. CPG 230, as guidance, says testing would be conducted by staff and teams independent of those with operational responsibility for controls.
    Evidence a web test can offer
    A dated report about one application's controls on the build tested: the scope as tested, findings, a CVSS v3.1 score and a business-risk rating for each rated finding, negative results, and test cases that could not run.
    What sits elsewhere, and who decides
    Whether the frequency is commensurate with the materiality of the risk, and whether a gap was rectified in a timely manner, are the entity's judgements. Controls outside the test's scope, such as backups, networks and staff devices, need their own testing.
    Limits of this row
    Read-only row: it logs test plans, report locations and dates, and sends no requests.
  15. E2. Remediation and the operational risk profile
    CPS 230 text, by paragraph number
    Paragraph 30: remediate material weaknesses in operational risk management, including control gaps, weaknesses and failures; the remediation must be supported by clear accountabilities and assurance and address the root causes of weaknesses in a timely manner; identified control gaps, weaknesses and failures are included in the operational risk profile until remediated.
    What to map or record
    Each finding not marked fixed: its owner, its target date, the root cause your team identified, and the dates it was entered in the operational risk profile and, later, removed from it.
    Evidence a web test can offer
    After the one consolidated retest, a Clever Ops report marks each finding fixed, partially fixed, still reproducible or not retested, on an identified build.
    What sits elsewhere, and who decides
    Whether a finding is a material weakness, and when it counts as remediated in the risk profile, are the entity's decisions. A finding marked not retested has not been confirmed fixed.
    Limits of this row
    Read-only row: it tracks owners, target dates and retest statuses, and sends no requests.
  16. E3. Incidents, near misses and the notification clocks
    CPS 230 text, by paragraph number
    Paragraph 31: ensure that operational risk incidents and near misses are identified, escalated, recorded and addressed in a timely manner. Paragraph 32: notify APRA as soon as possible, and not later than 72 hours, after becoming aware of an operational risk incident that the entity determines to be likely to have a material financial impact or a material impact on its ability to maintain its critical operations. Paragraph 41: notify APRA as soon as possible, and not later than 24 hours after, if it has suffered a disruption to a critical operation outside tolerance.
    What to map or record
    Who decides whether an event involving the web channel is an operational risk incident or a near miss, who notifies APRA and which clock applies, and the date each decision was made.
    Evidence a web test can offer
    Nothing. A test finding is evidence about a weakness on a tested build. It is not an incident report or a notification.
    What sits elsewhere, and who decides
    Each determination is the entity's. Footnote 11 says a notification of an information security incident reported under CPS 234 does not need to be separately reported under the notification requirements of CPS 230. Where personal information is involved, the Notifiable Data Breaches scheme has its own guide in this library.
    Limits of this row
    Read-only row: it logs who decided what, and when, and sends no requests.
  17. E4. The agreement with each material service provider
    CPS 230 text, by paragraph number
    Paragraph 53: for all material arrangements, maintain a formal legally binding agreement that, at a minimum, (a) specifies the services and associated service levels; (b) sets out each party's rights, responsibilities and expectations, including in relation to the ownership of assets, ownership and control of data, dispute resolution, audit access, liability and indemnity; (c) includes provisions to ensure the entity can meet its legal and compliance obligations; (d) requires notification by the provider of its use of other material service providers it materially relies upon in providing the service to the entity through sub-contracting or other arrangements; (e) makes the provider responsible for liability for any sub-contractor's failure; (f) includes a force majeure provision indicating which parts of the contract would continue in the case of a force majeure event; and (g) includes termination provisions including, but not limited to, the right to terminate the arrangement in its entirety or parts of it, with a further termination requirement for RSE licensees. Paragraph 54: provisions allowing APRA access to documentation, data and any other information related to the service, an on-site visit to the provider, and the provider's agreement not to impede APRA in fulfilling its duties as prudential regulator.
    What to map or record
    For each material arrangement in row M7: where the agreement is filed, the clause that covers each item in paragraphs 53 and 54 or a note that it is missing, whether paragraph 57 is relied on and why, and the next review date.
    Evidence a web test can offer
    Nothing towards the clauses themselves. Where an agreement asks the provider for security testing evidence, a report and its retest status are material the provider could share, as far as the agreement allows.
    What sits elsewhere, and who decides
    Whether an agreement meets paragraphs 53 and 54 is for the entity and its own advisers. A supplier reads its agreement with the customer and asks its own adviser what the agreement requires of it.
    Limits of this row
    Read-only row: it works from the signed agreements and sends no requests.
  18. E5. Monitoring each material arrangement
    CPS 230 text, by paragraph number
    Paragraph 59: monitor and ensure that senior management receive reporting on material arrangements commensurate with the nature and usage of the service, including a regular assessment of (a) performance under the service agreement with reference to agreed service levels; (b) the effectiveness of controls to manage the risks associated with the use of the service provider; and (c) compliance of both parties with the service provider agreement.
    What to map or record
    For each material arrangement: the date of each regular assessment, the service levels met or missed, the evidence of control effectiveness received, the issues raised, and when they were reported to senior management. CPG 230, as guidance, says periodic reviews could include assessment of operational issues (including information security incidents and service disruptions), control effectiveness, information security capabilities and business continuity capabilities, strategic changes and comparisons to other offerings in the market.
    Evidence a web test can offer
    A provider's test report and retest status can be part of the evidence of control effectiveness, for the systems the report covers.
    What sits elsewhere, and who decides
    Whether the evidence shows effective controls for paragraph 59(b) is the entity's assessment.
    Limits of this row
    Read-only row: it logs each assessment and the reporting that followed, and sends no requests.
  19. E6. Supplier: answering an APRA-regulated customer
    CPS 230 text, by paragraph number
    The paragraphs in this map are addressed to the APRA-regulated entity. Paragraph 14: an APRA-regulated entity must not rely on a service provider unless it can ensure that in doing so it can continue to meet its prudential obligations in full and effectively manage the associated risks. Paragraph 52: before entering into or materially modifying a material arrangement, (a) undertake appropriate due diligence, including an appropriate selection process and an assessment of the provider's ability to provide the service on an ongoing basis; and (b) assess the financial and non-financial risks from reliance on the provider, including risks associated with geographic location or concentration of the provider or parties it relies on in providing the service.
    What to map or record
    The customer's exact request, whether a questionnaire, a due diligence request, a contract clause or an invitation into a BCP exercise; whether the customer says you are a material service provider; the evidence you sent for each item; how much of it the agreement allows you to pass on; and the date sent.
    Evidence a web test can offer
    Evidence with a date and a stated scope, covering the systems you operate for that customer, and a retest status against each finding. A Clever Ops report on a system Clever Ops built or hosts also says so.
    What sits elsewhere, and who decides
    Your customer judges whether the evidence is enough. As for what your agreement with that customer asks of you, read it and put the question to your own adviser.
    Limits of this row
    Read-only row: it pairs each item in a customer's request with a document and sends no requests.

CPS 230 at a glance: which instrument is in force, and what it aims at

CPS 230 is Prudential Standard CPS 230 Operational Risk Management.

The text in force sits in the legislative instrument named Banking, Insurance, Life Insurance, Health Insurance and Superannuation (prudential standard) determination No. 1 of 2026. In it, a delegate of the Australian Prudential Regulation Authority (APRA) revokes determination No. 2 of 2023, including the CPS 230 made under that determination as varied by variation No. 1 of 2024, and determines the standard under subsection 11AF(1) of the Banking Act 1959, subsection 32(1) of the Insurance Act 1973, subsection 230A(1) of the Life Insurance Act 1995, subsection 92(1) of the Private Health Insurance (Prudential Supervision) Act 2015 and subsection 34C(1) of the Superannuation Industry (Supervision) Act 1993. The determination is dated 23 April 2026.

On the Federal Register of Legislation, the instrument's details page gives title ID F2026L00475, shows it as In force, and records its type as Principal, As Made, with a registration date of 29 April 2026. Paragraph 6 of the standard says it commences on 1 July 2026, and APRA's own page for CPS 230 marks it as in force from that date.

Paragraph numbers in this guide are those of the instrument as made. The register is the place to look for any later version before a paragraph is relied on.

The standard's opening states its aim: to ensure that an APRA-regulated entity is resilient to operational risks and disruptions. It says an APRA-regulated entity must effectively manage its operational risks, maintain its critical operations through disruptions, and manage the risks arising from service providers, and that the entity's approach to operational risk must be appropriate to its size, business mix and complexity.

Paragraph 12 says an APRA-regulated entity must identify, assess and manage operational risks that may result from inadequate or failed internal processes or systems, the actions or inactions of people or external drivers and events.

That is a wider brief than information security, and a different one. The words penetration test, website and cyber appear nowhere in the text of CPS 230. Where it reaches information security, paragraph 24 hands over to CPS 234, which has its own guide in this library.

This guide treats a website or web application as one of the resources a critical operation may depend on, which is how it reads paragraph 26(b) below.

Two readers are in mind: the risk or technology lead at an APRA-regulated entity who has to map what its online channels depend on, and the owner or technical lead of a business that builds, hosts or runs something for such an entity and has been sent a CPS 230 questionnaire, a due diligence request or a contract clause.

Nothing here is legal advice, and the guide is no substitute for an audit.

The guide reports the instrument's text and leaves each decision about meeting it with the people and bodies the instrument names. Confirm with your own adviser, or with APRA, whether the standard covers your business.

Why there is a 2026 version, and the previous version's transition for existing contracts

The 2026 determination replaced an earlier CPS 230.

APRA's explanatory statement for the 2026 determination says the previous CPS 230 came into effect on 1 July 2025, and, in its account of APRA's consultation, that the relevant contractual requirements for pre-existing arrangements applied from the earlier of the next contract renewal date or 1 July 2026.

Paragraph 6 of the 2026 text says the standard commences on 1 July 2026, and the application and commencement paragraphs set no separate date for service provider arrangements that were already in place. The explanatory statement says the purpose of the instrument is to revoke the previous CPS 230 and replace it with the new CPS 230, and that the new CPS 230 introduces a limited carve out from certain contractual obligations for eligible material arrangements with non-traditional service providers, while preserving the broader operational risk management, continuity and service provider risk management requirements in the previous CPS 230.

CPG 230 says paragraph 57 of CPS 230 outlines the exemptions available for such arrangements; the service provider section below sets it out.

Paragraph numbers are worth checking against the instrument whenever you read guidance or older material. In Table 1 of APRA's practice guide CPG 230, which APRA's page marks as current beside the date 1 July 2026, the 72-hour notice of an operational risk incident is listed as paragraph 33 of CPS 230.

In the 2026 instrument as made, that notice is paragraph 32. This guide uses the instrument's own numbering throughout.

Who APRA CPS 230 applies to, and where its suppliers come in

Under paragraph 2, APRA CPS 230 applies to all APRA-regulated entities, defined as: (a) authorised deposit-taking institutions (ADIs), including foreign ADIs, and non-operating holding companies authorised under the Banking Act (authorised banking NOHCs); (b) general insurers, including Category C insurers, non-operating holding companies authorised under the Insurance Act (authorised insurance NOHCs) and parent entities of Level 2 insurance groups; (c) life companies, including friendly societies, eligible foreign life insurance companies (EFLICs) and registered NOHCs; (d) private health insurers registered under the PHIPS Act; and (e) registrable superannuation entity licensees (RSE licensees) under the SIS Act in respect of their business operations.

Paragraphs 3, 4 and 10 qualify that list. Paragraph 3 limits the obligations the standard imposes on, or in relation to, a foreign ADI, a Category C insurer and an EFLIC to the Australian branch operations of that entity.

Paragraph 4 has an APRA-regulated entity that is the Head of a group comply with a requirement of the standard in its capacity as an APRA-regulated entity, by ensuring that the requirement is applied appropriately throughout the group, including in relation to entities that are not APRA-regulated, and on a group basis.

Under paragraph 10, APRA may adjust or exclude a specific prudential requirement in the standard in relation to an APRA-regulated entity.

A business outside those classes is not on the list. It can still be asked about CPS 230 when it supplies an entity that is. Paragraph 34 defines critical operations as processes undertaken by an APRA-regulated entity or its service provider which, if disrupted beyond tolerance levels, would have a material adverse impact on its depositors, policyholders, beneficiaries or other customers, or its role in the financial system.

Paragraph 14 then says an APRA-regulated entity must not rely on a service provider unless it can ensure that in doing so it can continue to meet its prudential obligations in full and effectively manage the associated risks.

The duties quoted on this page are placed on the APRA-regulated entity, its Board or its internal audit function, and the powers quoted are APRA's. APRA's guidance puts the point plainly: CPG 230 says that where an entity uses a service provider, the entity still owns and is responsible for managing its risk.

A supplier may come across CPS 230 through the entity's due diligence, its contract terms or its monitoring. If you supply such an entity, read your agreement with that customer and put the question of what it asks of you to your own adviser.

If your business may itself be one of the entities paragraph 2 lists, or part of a group headed by one, confirm with your adviser or APRA whether the standard covers you.

CPS 230 operational risk: critical operations, tolerance levels and what they depend on

The business continuity paragraphs start with a register.

Paragraph 33 says an APRA-regulated entity must (a) define, identify and maintain a register of its critical operations; (b) take reasonable steps to minimise the likelihood and impact of disruptions to its critical operations; (c) maintain a credible BCP that sets out how it would maintain its critical operations within tolerance levels through disruptions, including disaster recovery planning for critical information assets; (d) activate its BCP if needed in the event of a disruption; and (e) return to normal operations promptly after a disruption is over.

Paragraph 35 sets a floor for that register. An APRA-regulated entity must, at a minimum, classify the following business operations as critical operations, unless it can justify otherwise: for an ADI, payments, deposit-taking and management, custody, settlements and clearing; for an insurer (general, life, private health), claims processing; for an RSE licensee, investment management and fund administration; and for all APRA-regulated entities, customer enquiries and the systems and infrastructure needed to support critical operations.

Paragraph 36 says APRA may require an APRA-regulated entity, or a class of APRA-regulated entities, to classify a business operation as a critical operation.

For each critical operation, paragraph 37 says the entity must establish tolerance levels for (a) the maximum period of time the entity would tolerate a disruption to the operation; (b) the maximum extent of data loss the entity would accept as a result of a disruption; and (c) minimum service levels the entity would maintain while operating under alternative arrangements during a disruption.

Under paragraph 21(b), the Board must approve the BCP and those tolerance levels, review the results of testing and oversee the execution of any findings.

Where a website enters the picture

Paragraph 26(b) does not name a website, but the resources it lists include technology and service providers. 1 min read

The business continuity testing paragraphs

Paragraphs 42 to 45 sit under the heading Testing and review, and they deal with the business continuity plan. 1 min read

Controls, technology risk and the hand-over to CPS 234

Paragraph 23 says an APRA-regulated entity must manage its full range of operational risks, including but not limited to legal risk, regulatory risk, compliance risk, conduct risk, technology risk, data risk and change management risk.

Paragraph 24 turns to technology itself: the entity must maintain appropriate and sound information and information technology (IT) capability to meet its current and projected business requirements and to support its critical operations and risk management, and, in managing technology risks, it must monitor the age and health of its information assets and meet the requirements for information security in Prudential Standard CPS 234 Information Security.

Paragraphs 29 and 30, under the heading Operational risk controls, deal with testing and remediation. Paragraph 29 says an APRA-regulated entity must regularly monitor, review and test controls for design and operating effectiveness, the frequency of which must be commensurate with the materiality of the risks being controlled, and that the results of testing must be reported to senior management and any gaps or deficiencies in the control environment must be rectified in a timely manner.

Paragraph 30 says it must remediate material weaknesses in its operational risk management, including control gaps, weaknesses and failures; that the remediation must be supported by clear accountabilities and assurance and address the root causes of weaknesses in a timely manner; and that identified control gaps, weaknesses and failures stay in its operational risk profile until they are remediated.

CPG 230 adds, as guidance, that testing would be conducted by staff and teams independent of those with operational responsibility for controls. Among the things it says entities could consider when they monitor, review and test the effectiveness of controls are testing controls for material risks more frequently than for less material risks, and capturing all controls, including those owned by related parties and service providers.

Information security has a standard of its own. Paragraph 24 sends it to CPS 234, and this library's CPS 234 guide sets out that standard's testing paragraphs and what a web application test can add to them. This page stays with the operational side: what the web channel supports, what it depends on, and what happens when a piece of it fails.

CPS 230 service provider management: the policy, the register and the agreement

Paragraph 46 says an APRA-regulated entity must maintain a comprehensive service provider management policy, covering how the entity will identify material service providers and manage service provider arrangements, including the management of material risks associated with the arrangements.

Paragraph 47 says the policy must include (a) the entity's approach to entering into, monitoring, substituting and exiting agreements with material service providers; (b) its approach to managing the risks associated with material service providers; and (c) its approach to managing the risks associated with any fourth parties that material service providers rely on to deliver a critical operation to the APRA-regulated entity.

Footnote 14 says a fourth party is a party that a service provider relies on in delivering services to an APRA-regulated entity.

Paragraph 48 says an APRA-regulated entity must identify and maintain a register of its material service providers and manage the material risks associated with using them. Material service providers, it says, are those on which the entity relies to undertake a critical operation or that expose it to material operational risk.

Footnote 15 adds that a material service provider may be a third party, related party or connected entity, and that a service provider may be identified as material as a result of an individual arrangement or multiple arrangements with an APRA-regulated entity.

Paragraph 49 lists services whose providers an APRA-regulated entity must, at a minimum, classify as material service providers, unless it can justify otherwise: for an ADI, credit assessment, funding and liquidity management and mortgage brokerage; for an insurer (general, life, private health), underwriting, claims management, insurance brokerage and reinsurance; for an RSE licensee, fund administration, custodial services, investment management and arrangements with promoters and financial planners; and for all APRA-regulated entities, risk management, core technology services and internal audit.

Paragraph 50 has the entity submit its register to APRA on an annual basis, and paragraph 51 says APRA may require an entity, or a class of entities, to classify a service provider, type of service provider or service provider arrangement as material.

Not every supplier is material. CPG 230 says, as guidance, that for the purposes of the register CPS 230 does not intend to capture arm's length transactions or intermediation unless they meet the criteria under paragraph 48. The classification itself belongs to the entity.

Due diligence and the formal agreement

Paragraph 52 says that before entering into or materially modifying a material arrangement, an APRA-regulated entity must (a) undertake appropriate due diligence, including an appropriate selection process and an assessment of the ability of the service provider to provide the service on an ongoing basis; and (b) assess the financial and non-financial risks from reliance on the service provider, including risks associated with geographic location or concentration of the service provider(s) or parties the service provider relies on in providing the service. 1 min read

APRA's access, an orderly exit and the 2026 carve out

Paragraph 54 says the formal agreement must also include provisions that allow APRA access to documentation, data and any other information related to the provision of the service, allow APRA the right to conduct an on-site visit to the service provider, and ensure the service provider agrees not to impede APRA in fulfilling its duties as prudential regulator. 2 min read

Monitoring, notice to APRA and internal audit

Paragraph 59 says an APRA-regulated entity must monitor and ensure that senior management receive reporting on material arrangements commensurate with the nature and usage of the service, and that this monitoring must include a regular assessment of (a) performance under the service agreement with reference to agreed service levels; (b) the effectiveness of controls to manage the risks associated with the use of the service provider; and (c) compliance of both parties with the service provider agreement. 1 min read

Reading these paragraphs as a supplier

The service provider paragraphs place their duties on the APRA-regulated entity, not on its suppliers. 1 min read

What a website or web application test can evidence under CPS 230

Nothing in the text of CPS 230 names a website test.

Where a website or web application supports a critical operation, a test of it is evidence about that one resource as it stood: one build, one test window, the accounts and roles the tester held, and a written scope. Set beside the paragraphs above, a report can contribute these items and nothing beyond them.

  • Towards paragraph 26(b): the scope as tested and, from an external check, an inventory of the hostnames that answer on the domain, with where each one points and who appears to run it. That helps fill rows M3 to M7 of the map. It does not decide which operations are critical or what each one depends on.
  • Towards paragraph 29: a dated record of how particular controls on one application behaved, such as sign-in, sessions, access between customers and between roles, and the application's side of a payment. Whether the testing is frequent enough for the materiality of the risk is the entity's judgement.
  • Towards paragraph 30: each finding's ratings and its status after the retest, which show which gaps stayed open on the tested build. Whether a finding is a material weakness, and when it leaves the operational risk profile, is decided inside the entity.
  • Towards paragraph 59: where a supplier runs the application, its report and retest status can be part of what the entity weighs in its regular assessment of the effectiveness of controls, for the systems the report covers. The entity makes that assessment.
  • Towards the scenario analysis in paragraph 26(c) and the BCP testing in paragraphs 42 and 43: the facts a tabletop exercise needs about a real weakness, such as the affected routes, the build and the criteria a fix has to pass at the retest.

What it cannot speak for

A web application test says nothing about whether a critical operation can be kept within its tolerance levels; about recovery times, whether backups and restores work, or disaster recovery; about the BCP and its annual exercise; about whether a provider is material or an agreement meets paragraphs 53 and 54; or about controls its scope did not reach, among them office networks, staff devices and the platforms of vendors the application relies on.

It does not test behaviour under load. A finding is not an operational risk incident report, and a test is not a notification to APRA.

A report also makes no finding that an entity or a supplier meets CPS 230; the methodology guide sets out why a report describes what was tested and does not issue a certificate.

Filling in the dependency map and evidence register

The table on this page has three parts.

Rows M1 to M8 map each critical operation the web channel supports to the public website, the customer app or portal, identity, payments, and the service providers and fourth parties behind them. Rows T1 to T5 are tolerance scenarios to rehearse on paper. Rows E1 to E6 are the evidence register.

An APRA-regulated entity can use the table to track what its online channels rely on. A supplier can fill rows M3 to M8 for the systems it runs, and rows E4 and E6 for each regulated customer.

  • Start with rows M1 and M2, copying names and tolerance levels from the entity's own register and BCP. A supplier gets these from its customer, or leaves them blank and says so.
  • Fill rows M3 to M7 from records you already hold: DNS and hosting accounts, form settings, the app's architecture notes and your supplier list. None of those rows needs anyone to open a screen that shows a password, key or secret.
  • Row M8 alone sends requests: three page loads, one each, with nothing signed in to and no form submitted, on a site and app your business runs or holds written authority to test, using a staging copy where one exists. Host names are all it records, and its full limits sit on the row.
  • Run rows T1 to T5 as tabletop exercises. Each outage they describe is invented and talked through, never caused: no scenario row has anyone switch off, overload or interrupt a live service.
  • Before testing or changing anything in a system a vendor runs, read that vendor's terms, and put any question they leave open to the vendor. Nothing on this page states what a vendor permits.
  • Put real dates and owners in every evidence row, and keep the evidence for what passed as well as what failed.
  • If someone else's real data appears while you work through a row, halt at that point: write down where and when you saw it, leave everything else unopened, and tell the person who handles privacy and incidents in your business.

Who decides what the evidence shows

The standard names the deciders.

Paragraph 19 says the Board of an APRA-regulated entity is ultimately accountable for oversight of the entity's operational risk management, including business continuity and the management of service provider arrangements. Under paragraph 21, the Board must approve the BCP and tolerance levels for disruptions to critical operations, review the results of testing and oversee the execution of any findings, and approve the service provider management policy and review risk and performance reporting on material service providers.

Paragraph 29 sends the results of control testing to senior management, paragraph 40 has any failure to meet tolerance levels reported to the Board together with a remediation plan, and paragraphs 45 and 61 give internal audit its reviews.

APRA keeps powers of its own. Paragraph 10 says it may adjust or exclude a specific prudential requirement in relation to an APRA-regulated entity, and paragraph 18 lists what APRA may do where it considers that an entity's operational risk management has material weaknesses, among them requiring an independent review of it and imposing conditions on the entity's licence. A tester makes none of these decisions.

A supplier's position turns on its customer's decisions. The regulated customer decides whether you are material, subject to APRA's power under paragraph 51, what it needs from you and whether your evidence is enough. A report lets you answer with dated facts.

Whether those facts satisfy the customer is the customer's call. What your agreement with the customer requires of you is a question to read the agreement for and to put to your own adviser.

Where Clever Ops testing sits in CPS 230 work, and its limits

Three Clever Ops services, each testing something customers reach from a browser or through an API, correspond to rows of the map.

The website security check is an outside view of the public site, without signing in, and its hostname inventory and its certificate and DNS results feed rows M3, M7 and T1. The WordPress security audit fits a public site built on WordPress. The web application and API penetration test fits a custom app or portal that customers or staff sign in to, and its results can feed rows M4 to M6, T2 to T4, E1, E2, E5 and E6.

Each report sets out the scope as tested; its findings with redacted evidence, and for every rated finding a CVSS v3.1 score and a business-risk rating; negative results for the build and accounts tested; and the test cases that could not run.

Each test comes with one consolidated retest, and the updated report then gives every finding one status: fixed, partially fixed, still reproducible or not retested.

None of the three is an operational resilience assessment. Their scopes, as their own pages describe them, cover the security of a site, an application or an API; measuring recovery times, testing backups and restores, running business continuity exercises, classifying service providers and reviewing agreements are not among them.

Load, volumetric and denial-of-service testing sit outside every Clever Ops service. Clever Ops has not been certified or accredited under any security scheme. A Clever Ops report is evidence for a decision that the entity, its Board, its internal audit function or APRA makes, and it is not a certificate, a CPS 230 assessment or a compliance opinion.

Where an entity, an agreement or a tender requires a CREST-accredited provider, that work goes to a provider holding the accreditation, because Clever Ops is not CREST accredited.

Independence matters here too. In its section on maintaining effective controls, CPG 230 says, as guidance, that testing would be conducted by staff and teams independent of those with operational responsibility for controls. For an application that Clever Ops itself built or hosts, the engagement is labelled a security review instead of an independent test, and the report records which business built the system and which one hosts it.

An entity or customer that wants a tester untouched by both the build and the hosting should engage a provider that had no hand in either.

Clever Ops does not test a vendor's own product or platform, such as an identity or payment provider's service, and work on internal networks, staff devices or red teaming is referred to a network or red team specialist.

Frequently asked questions

The version in force is the one made by determination No. 1 of 2026. Its paragraph 6 says it commences on 1 July 2026, and its text sets no separate date for arrangements already in place. It replaced an earlier version: APRA's explanatory statement for the 2026 determination says the previous CPS 230 took effect on 1 July 2025, with the relevant contractual requirements for pre-existing arrangements applying from the earlier of the next contract renewal date or 1 July 2026. Check the Federal Register of Legislation for any later version before relying on a paragraph.

CPS 230 does not name one: the words penetration test, website and cyber are absent from its text. Paragraph 29 says an APRA-regulated entity must regularly monitor, review and test controls for design and operating effectiveness, with a frequency commensurate with the materiality of the risks being controlled, and paragraph 42 says it must have a systematic testing program for its BCP that covers all critical operations and includes an annual business continuity exercise. Paragraph 24 sends information security to CPS 234, which has its own testing paragraphs and its own guide in this library. Which tests an entity runs to meet these paragraphs, and on which systems, is for the entity to settle, with its own advisers where it needs them.

That is the bank's classification to make. Paragraph 48 says material service providers are those on which the entity relies to undertake a critical operation or that expose it to material operational risk. Paragraph 49 says an APRA-regulated entity must, at a minimum, classify a provider of the services it lists as a material service provider, unless it can justify otherwise, and the services it lists for all APRA-regulated entities are risk management, core technology services and internal audit. Paragraph 51 says APRA may require an entity to classify a service provider as material. CPG 230 says, as guidance, that CPS 230 does not intend to capture arm's length transactions or intermediation unless they meet the criteria under paragraph 48. The bank can tell you how it has classified you. For what your agreement with the bank asks of you, read the agreement and take the question to your own adviser.

Paragraph 53 says that for all material arrangements, the entity must maintain a formal legally binding agreement that, at a minimum, specifies the services and associated service levels; sets out each party's rights, responsibilities and expectations, including in relation to the ownership of assets, ownership and control of data, dispute resolution, audit access, liability and indemnity; includes provisions to ensure the entity can meet its legal and compliance obligations; requires the provider to notify its use of other material service providers it materially relies upon in providing the service to the entity through sub-contracting or other arrangements; makes the provider responsible for liability for any failure by a sub-contractor; includes a force majeure provision indicating which parts of the contract would continue in a force majeure event; and includes termination provisions including, but not limited to, the right to terminate the arrangement in its entirety or parts of it, with a further termination requirement for RSE licensees. Paragraph 54 adds provisions allowing APRA access to documentation, data and any other information related to the service, the right to an on-site visit to the provider, and the provider's agreement not to impede APRA in fulfilling its duties as prudential regulator. Under paragraph 57, an entity need not comply with paragraphs 53, 54, 55(d), 59(a) and 59(c) for a material arrangement if the conditions in its (a) and (b) are met: the provider falls within a category listed in the Attachment, and the arrangement uses standardised terms or is not documented in a formal agreement. Whether a given agreement meets those paragraphs is for the entity and each party's own adviser.

CPS 230 deals with operational risk, business continuity and service providers, and CPS 234, as its title says, deals with information security. Paragraph 24 of CPS 230 says that in managing technology risks, an APRA-regulated entity must monitor the age and health of its information assets and meet the requirements for information security in CPS 234. Footnote 11 says a notification of an information security incident reported under CPS 234 does not need to be separately reported under the notification requirements of CPS 230. The CPS 234 guide in this library covers that standard's testing paragraphs and what a web application test can add to them.

No. A report records what was tested on one application, build and scope, and in CPS 230 terms it is one input to the entity's control testing under paragraph 29 and its remediation record under paragraph 30. The entity, its Board and internal audit function, and APRA are the ones who weigh whether an entity meets the standard. A supplier's evidence is judged sufficient or not by its regulated customer. A Clever Ops report sets down what was in scope, what turned up and how the retest went; it is neither a certificate nor a compliance opinion.

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.