APRA's Prudential Standard CPS 230 Operational Risk Management took effect on 1 July 2025. It requires regulated entities to identify critical operations, set board-approved tolerance levels, test controls and actively manage material service providers. AI helps with continuous control monitoring, evidence assembly, vendor oversight and incident timelines. Accountable people keep every judgement on tolerances and reporting.
This guide is for risk, operations, technology and finance leads at banks, mutuals, insurers and super funds, and for the technology and services firms that supply them. It sets out where automation earns its place under CPS 230 and where a person has to own the decision.
Key takeaways
- CPS 230 moves operational risk from point-in-time paperwork to continuous, evidenced control.
- AI suits the high-volume work: monitoring control logs, assembling evidence packs, tracking vendor SLAs and renewals, and building incident timelines.
- Setting tolerance levels, deciding what is critical and deciding whether to notify APRA stay with accountable people and the board.
- An AI system used in or near a critical operation is itself an operational risk, and its provider may be a material service provider.
- Service providers to regulated entities feel CPS 230 through contracts, and clean evidence of their own controls is a selling point.
What does APRA CPS 230 require?
CPS 230 replaced older standards on outsourcing (CPS 231) and business continuity (CPS 232) with a single regime. It applies to authorised deposit-taking institutions, general insurers, life companies, private health insurers and RSE licensees.
- A sound operational risk management framework, with controls identified, monitored and tested, beyond being documented once.
- Identification of critical operations and board-approved tolerance levels for how long they can be disrupted, how much data can be lost and how quickly they must recover.
- A service provider management policy, formal agreements and active monitoring of the parties you rely on.
- A register of material service providers, linking each to the critical operations it supports.
How is CPS 230 different from the old outsourcing and BCM standards?
The simplest way to read CPS 230 is as a move from point-in-time paperwork to continuous, evidenced control. Under the older standards, much of the obligation could be met with a policy, an annual review and a contract on file. CPS 230 expects you to know your critical operations, test that you can keep them running inside tolerance, and watch the providers underneath them.
| Old outsourcing / BCM standards | CPS 230 from 1 July 2025 | Impact | |
|---|---|---|---|
| Focus | Outsourcing and continuity in silos | Single operational risk framework | Unified |
| Controls | Documented, reviewed periodically | Monitored and tested for effectiveness | Continuous |
| Critical operations | Loosely defined | Identified with board-approved tolerances | Explicit |
| Service providers | Material outsourcing only | Broader material service provider register | Wider |
| Board role | Oversight in principle | Approves tolerances and BCP directly | Accountable |
The consequence is a large increase in evidence that must be produced and kept current: control testing, incident records, provider performance data and tolerance breach analysis, captured continuously. That is high-volume, repetitive work where AI can carry real load while judgement stays human.
Where does AI help with CPS 230 compliance?
CPS 230 wants continuous monitoring, consistent record-keeping and fast synthesis of operational data. Those suit automation, as long as accountable people review the output.
- IngestControl logs, incident tickets, vendor SLAs
- MonitorFlag control failures and tolerance drift
- EvidenceAssemble records for each critical operation
- EscalateRoute exceptions to a named risk owner
- Continuous control monitoring. AI watches control logs, incident systems and change records and surfaces where a control is failing or drifting before it becomes a reportable event. The risk owner decides what it means.
- Evidence assembly. AI pulls the logs, test results and incident history for a critical operation into an evidence pack for a person to review and sign.
- Service provider oversight. AI tracks SLA performance, renewal dates and reported incidents across a large vendor base and flags where a material provider is slipping.
- Incident synthesis. When something goes wrong, AI assembles the timeline from logs and tickets so the review starts from facts. Root cause and any notification decision stay with people.
Incident synthesis inside a client's own environment is the problem we built RootCauseAI, our own product, to solve. The lesson from building it applies here: the value is getting a defensible timeline in front of an expert quickly, with the expert deciding what the incident means.
Who sets CPS 230 critical operations and tolerance levels?
People do. A critical operation is a process that, if disrupted beyond tolerance, would have a material adverse impact on depositors, policyholders, beneficiaries or other customers, or on the entity's role in the financial system. Tolerance levels cover maximum unavailability, maximum data loss, recovery time and the minimum service levels needed to keep operating.
The board approves the business continuity plan and tolerance levels, reviews testing results and oversees the response to findings. AI can map dependencies, model scenarios and keep evidence current. The decision about what is critical and how much disruption is acceptable is a governance decision for accountable people.
What are you trying to do?
- Monitor controls and gather evidence continuouslyGood fit for AI, with human sign-off
- Track vendor SLAs and renewal dates at scaleGood fit for AI
- Set tolerance levels for a critical operationBoard and management decision
- Decide whether a disruption must be notified to APRAAccountable person decides
What are the CPS 230 material service provider deadlines?
A service provider is material if the entity relies on it to undertake a critical operation, or if the arrangement exposes the entity to material operational risk. Regulated entities must keep a register of these providers, link each to the critical operations it supports and be able to submit it to APRA.
- 1 Jul 2025CPS 230 in forceOperational risk and business continuity obligations commence
- 1 Oct 2025First register dueMaterial service provider registers submitted to APRA
- 1 Jul 2026Existing contracts capturedPre-existing agreements covered from the earlier of renewal or this date
- OngoingKeep the register currentCheck the current CPS 230 and CPG 230 text on APRA's site
For existing contracts, CPS 230's service provider requirements apply from the earlier of the next renewal or 1 July 2026. APRA has also refined the standard and its Prudential Practice Guide CPG 230 since commencement, including on contractual requirements for some arrangements. If your register and contract remediation work was built on an earlier draft, review it against the current text on APRA's website.
This is where AI helps with the mechanical load: reading a large contract portfolio to flag which agreements touch critical operations and which lack the clauses CPS 230 expects. A governed AI contract review workflow can triage a vendor book far faster than a manual pass, leaving legal and procurement to make the judgement calls.
How does CPS 230 affect service providers to APRA-regulated entities?
You do not need to be APRA-regulated to feel CPS 230. If you supply a regulated entity and sit near one of its critical operations, your client must bring you inside its service provider management regime: stronger contract terms, rights to monitor your performance, notification obligations and evidence that you can support the tolerances your client has committed to.
For a fintech, claims platform, administration provider or data business, that is a cost and an opportunity. The cost is contractual and assurance overhead. The opportunity is that a provider who arrives at procurement with clean evidence of its controls, incident history and continuity testing is easier to onboard than one who cannot.
- Continuous control monitoring vs periodic manual reviewFewer surprises
- Evidence assembly per critical operationAudit-ready
- Vendor SLA and renewal tracking across the registerFewer missed dates
- Incident timelines for reviewsFaster start
- Human judgement retained on tolerances and reportingAccountable
These outcomes are deliberately qualitative. CPS 230 is about the reliability of your operations, and any quantified benefit depends on your starting point. Measure your own baseline before putting a number in a board pack.
Where can AI add risk under CPS 230?
AI cannot own accountability. Named individuals and the board are accountable for operational risk, tolerance levels and decisions after a disruption. A system that assembles evidence or flags a control failure has not made a decision. Any process where "the model said it was fine" replaces a sign-off has misread the standard.
AI becomes part of your operational risk surface. If you deploy AI into a critical operation, for example to triage incidents or monitor controls, it must be understood, tested, monitored and included in continuity planning like any other dependency. If a third party provides it, that provider may belong on your material service provider register.
Data handling matters. Sending operational logs and customer-adjacent information to an AI tool without knowing where data goes, how long it is kept and whether it trains a shared model creates privacy and confidentiality exposure. Our article on how the wrong AI tools leak business data covers the failure modes, and our data sovereignty guide explains why regulated entities often keep AI on their own infrastructure in Australia.
How does CPS 230 AI fit into wider AI governance?
AI for CPS 230 work sits inside how you govern AI generally: who approves a use case, what data it can touch, who signs off outputs and how you evidence all of that. Our AI governance framework and AI vendor selection questions are good starting points.
For security controls, the 50-point AI security checklist gives a baseline to test against before any AI system goes near a critical operation. CPS 230 also overlaps with the incident reporting obligations in the Cyber Security Act 2024, which apply to many of the same entities. Our guide to financial services AI compliance under APRA and ASIC covers the broader picture.
What should you do this quarter?
- Reconcile your register against the current standard. Confirm your material service provider register reflects the current CPS 230 and CPG 230 text.
- Triage your contract portfolio. Use a governed AI review to flag which agreements touch critical operations and which lack expected clauses.
- Map your AI systems as dependencies. Any AI tool near a critical operation belongs in your operational risk thinking, and its provider may be material.
- Write the sign-off rule down. Every AI-generated flag, evidence pack or incident timeline is verified by a named person before it informs a decision or reaches the board or APRA.
- If you are a provider, get audit-ready. Assemble evidence of your controls, incidents and continuity testing.
How Solve8 can help
Our consultants and engineers build CPS 230 evidence and monitoring workflows that run on your infrastructure in Australia, with a named person approving every output before it reaches the board or APRA. See our financial services and custom AI development pages, or talk to us.
This article is general information, not legal or prudential advice. Confirm your obligations against APRA's current CPS 230 and CPG 230 and qualified advice.
