PCI DSS integration means designing your integration layer so that as few systems as possible touch card data, and so that those which do are isolated, logged and tested. The standard’s own scoping guidance says the same: keep account data in as few places as possible and segment the rest. Get that right and compliance becomes an engineering property of the architecture, not an annual audit scramble. Our enterprise integration services are built around this principle: integration, APIs and events designed with governance in from the start.
This guide is for CTOs, CEOs and integration managers who own estates where payments flow through many systems: banks, healthcare providers, telecoms, logistics and retail.
Key Takeaways
- PCI DSS covers every system that stores, processes or transmits card data, plus anything that could affect their security. Every integration you add can widen that scope.
- The cheapest control is a smaller scope: eliminate card data you do not need, consolidate what you do, and segment the rest, then prove the segmentation by penetration testing.
- In the cloud, PCI DSS duties are shared between provider and customer. Your integration platform, APIs and configuration remain yours to evidence.
- AI does not change the rules. AI components that touch payment data need human oversight, logging and access limited to protected data.
- Version 4.0.1 is the current standard. Treat it as a continuous programme (automated log review, script control, regular segmentation tests), not a yearly event.
Table of Contents
- What PCI DSS covers, and who is responsible
- The current standard in brief
- Why integration is where scope grows
- A PCI compliant architecture, step by step
- Where the integration platform fits
- PCI DSS for cloud
- Keeping AI inside PCI DSS
- Payment pages and script risk
- Who this matters most to
- Our perspective
- Bringing it into your estate
- FAQ
- Sources
What PCI DSS covers, and who is responsible
The Payment Card Industry Data Security Standard is a baseline of technical and operational requirements for protecting payment account data. According to the PCI Security Standards Council, it is intended for entities that store, process or transmit cardholder data or sensitive authentication data, or that could affect the security of the cardholder data environment (CDE). That includes merchants, processors, acquirers, issuers and service providers.
Size and geography do not exempt you. AWS’s security reference architecture notes that the standard applies regardless of transaction volume, and that it was created jointly by Visa, Mastercard, American Express, Discover and JCB.
Who decides whether you must validate compliance? Not the standard itself. The PCI SSC says that is down to the organisations running compliance programmes: the card brands, your acquirer or other entities. For a UK enterprise, your acquiring bank is normally the first conversation. Responsibility, though, sits with you. Outsourcing a function does not outsource accountability, a point that matters more with every integration partner you add.
The current standard in brief
PCI DSS v4.0.1 is dated June 2024, following v4.0 in March 2022. It organises controls into twelve principal requirements, as summarised by AWS:
- Install and maintain network security controls
- Apply secure configurations
- Protect stored account data
- Protect cardholder data with strong cryptography in transmission
- Protect from malicious software
- Develop and maintain secure systems and software
- Restrict access by business need to know
- Identify users and authenticate access
- Restrict physical access
- Log and monitor all access
- Test security regularly
- Support information security with organisational policies
For integration leaders, several changes carry architectural weight:
- Multi-factor authentication. Requirement 8.4.2 now applies to all non-console access into the CDE, from any location, not only remote access, according to SecurityMetrics. Version 4.0.1 clarified that it does not apply to accounts that authenticate exclusively with phishing-resistant factors, per Checkout.com.
- Automated log review. Requirement 10.4.1.1 makes automated review (for example, SIEM tooling) mandatory. Manual review is no longer enough.
- Payment page scripts. Requirement 6.4.3 requires scripts on payment pages to be documented, tracked and authorised. Requirement 11.6.1 requires monitoring for malicious scripts, with a seven-day default frequency unless a risk analysis justifies another. Both target web-skimming.
- Patching. The 30-day window applies to critical vulnerabilities specifically, as the 4.0.1 clarification confirmed.
For merchants on the simplest self-assessment, the revised SAQ A (effective 31 March 2025) removed 6.4.3 and 11.6.1 and replaced them with an eligibility criterion: the merchant confirms its site is not susceptible to script-based attacks. Requirement 11.6.1 still applies to SAQ A-EP, SAQ D for merchants, SAQ D for service providers and onsite assessments. Most enterprises with custom integration sit in the heavier categories, so assume the script controls apply to you and confirm the category with your acquirer or assessor.
Why integration is where scope grows
PCI DSS applies to the CDE and to system components, people and processes that could affect the security of cardholder data. The PCI SSC’s 2024 scoping guidance also includes components with unrestricted connectivity to those systems.
That last clause is the integration problem. A shared message broker, a flat network, an API gateway that routes both orders and payments, a reporting database that receives a copy of transaction data: each can pull otherwise unrelated systems into scope. In a legacy-heavy estate built over fifteen years, the card data often sits in places nobody mapped. The same guidance warns that restricting account data to as few locations as possible may require an organisation to reengineer long-standing business practices. That is an architecture task, and it is cheaper to do it once than to audit every system annually.
A PCI compliant architecture, step by step
1. Map where card data actually flows
Start with data flows, not network diagrams. For every channel (web, call centre, point of sale, partner API, batch file), trace where the primary account number enters, where it is stored, which services pass it on and where it leaves. Include logs, queues, dead-letter stores, analytics extracts and backups, which are the places card data hides. Your integration platform’s message tracing and API inventory are the fastest sources of truth here.
The PCI SSC published scoping and segmentation guidance in December 2016 for exactly this exercise, and refreshed it for modern architectures in 2024.
2. Eliminate and consolidate card data
The most effective control is not holding the data. Ask of every system: does it need the full number, or would a token do? Where a business process only needs to recognise a card, not use it, replace the number at the edge so that downstream services never see it. Where data must be kept, consolidate it into one hardened store rather than copies across applications.
This is also where integration design matters. If an orchestration flow carries the payment payload through five services, all five are in scope. If it passes a token and calls the payment service only at the final step, one is.

3. Segment the CDE and prove it
To be out of scope, a component must be properly segmented so that it could not affect account-data security even if compromised. The guidance says segmentation must be verified by penetration testing that confirms connections to the CDE are not possible.
The testing cadence is specific. Under Requirement 11.4.5, most organisations test segmentation at least every twelve months and after any change to segmentation controls. Service providers test every six months under Requirement 11.4.6, as one engineering write-up explains. Build the test into your change process: a new integration route is a segmentation change.
The PCI SSC’s then Chief Technology Officer, Troy Leach, put the benefit plainly in the 2016 announcement: good segmentation lets an organisation focus on a limited number of assets and should reduce the effort to comply.
4. Control the interfaces
Every API, event topic and file transfer that crosses the CDE boundary needs three things: authentication that meets Requirement 8 (including MFA where the interface is administrative), strong cryptography in transit under Requirement 4, and a logged, reviewable record under Requirement 10. Treat boundary interfaces as products with owners, schemas and versioning, not as point-to-point plumbing.
5. Automate evidence
Continuous compliance means the architecture produces its own evidence: configuration baselines checked automatically, logs shipped to a monitored store, segmentation tests triggered by change, and certificates and keys rotated on schedule. With automated log review now mandatory, this is no longer optional.
Where the integration platform fits
Most enterprise payment flows run through an integration platform or API gateway. Choosing and configuring it is a PCI decision as much as an engineering one. We work with Boomi, MuleSoft, TIBCO and Solace, and the same design questions apply whichever is in place:
| Design question | Why it matters for PCI DSS |
|---|---|
| Does the payload carry a token or the card number? | Decides how many downstream services are in scope |
| Are payment flows on separate runtimes, environments or topics? | Supports segmentation and keeps non-payment traffic out of scope |
| Who can deploy and administer the platform? | Requirements 7 and 8: need-to-know access and MFA |
| Are message logs and payload tracing storing card data? | Requirement 3: stored data must be protected, and logs are stored data |
| Is every flow logged and reviewed automatically? | Requirement 10, including automated review |
Two patterns recur. Iconnect-style legacy modernisation tends to uncover payment data in message archives and monitoring consoles, and event platforms such as Solace need topic-level separation so that payment events are not readable by general subscribers. Neither is a reason to avoid the platform: they are design choices made early. You can see how we approach the platforms on our integration platforms page.
PCI DSS for cloud
Moving the CDE to the cloud does not move the obligation. In a cloud environment, PCI DSS responsibilities are shared between the provider and the customer. AWS, for example, provides its Report on Compliance and a shared responsibility matrix through AWS Artifact, to help customers meet Requirement 12.9, as its guidance describes.
The PCI SSC Cloud Computing Guidelines define which services and components the provider is responsible for (IaaS, PaaS or SaaS), and the scope of any extra managed security services it provides. A layered example is VMware Cloud on AWS: AWS is responsible for facilities and hardware, VMware for the software-defined data centre, and the customer for deployment, configuration, virtual machines and data.
The practical rule: the provider secures the platform, you secure what you build on it. Identity, network rules, encryption configuration, integration flows and application code remain yours. Write the split down per service, per requirement, and keep it with your evidence.

Third-party providers
Every payment gateway, hosting provider and integration partner is a third-party service provider (TPSP). Requirement 12.8 obliges you to manage and monitor them. A new PCI SSC information supplement on understanding service providers reinforces this. Under Requirement 12.9, a TPSP acknowledges its responsibility in writing and supports your requests for information. Where it meets requirements on your behalf or can affect your environment, it should give you evidence that it meets those applicable to its service. Keep a matrix of which requirement each provider covers, and renew the evidence each year.
Keeping AI inside PCI DSS
AI is entering payment-adjacent workflows: fraud scoring, customer service assistants, code generation, reconciliation. The PCI SSC’s AI principles are clear that using AI does not remove or bypass PCI requirements. AI systems should not be trusted with high-impact secrets or unprotected sensitive data, and should have human oversight, logging, monitoring and access only to suitably protected account data.
Three consequences for architecture:
- Keep card data away from models. Tokenise or mask before any data reaches an AI component, so the model never sees account numbers.
- Keep a human accountable. The principles call for a human in the loop with responsibility for the end-to-end process. AI can perform individual actions when building, testing and deploying software, but the whole pipeline should not be fully automated.
- Log the reasoning. Requirement 10 applies: actions by the AI should be logged and monitored, with a named person responsible and, where possible, enough logging to audit the prompts and reasoning behind outputs.
In October 2026 the PCI SSC published additional guidance on securing AI in payment environments, developed with the Global Executive Assessor Roundtable and its Board of Advisors. It is not mandatory, and where it differs from the PCI standards, the standards take precedence. We cover the engineering moves in 3 moves to keep AI PCI DSS compliant, and the governance side in our piece on AI risk management for enterprise leaders.
Payment pages and script risk
For any enterprise with a web checkout, the browser is part of the attack surface. Requirements 6.4.3 and 11.6.1 exist because third-party scripts on payment pages can skim card data without touching your servers. Keep an inventory of every script with a written justification and an authoriser, and monitor for change. The wider case for protecting online payments is set out in this piece on cybersecurity for e-commerce.
Who this matters most to
The cost of getting this wrong falls unevenly. Banks and insurers face it in core-systems integration, which we examine in banking systems integration. Telecoms and logistics firms take payments across many channels and partners. Healthcare providers combine payment data with patient data, so PCI DSS sits beside other regimes; our guide to HIPAA compliant AI controls covers the overlap for US-regulated work. In every case the pattern is the same: the more systems an integration connects, the more it matters that the connection is designed, not accreted.
Our perspective
We approach PCI DSS as an integration design problem. Most compliance cost in a large estate is not the twelve requirements themselves. It is the number of systems that the requirements land on, because nobody decided where card data should and should not travel.
Our teams are senior and nearshore, based in the UK and Portugal, with certified and audited practices for regulated environments. We design integration layers, APIs and event platforms with governance and accountability from the outset, so that scope, access and logging are architecture decisions taken early, not controls bolted on before an audit.
That matters most in complex, legacy-heavy estates, where generic teams struggle and where undocumented data flows are common. It matters equally for AI: a model joining a payment workflow should inherit the same segmentation, logging and human accountability as any other component.
Polarising Team
Bringing it into your estate
If you are planning a migration, a platform consolidation or an AI programme that touches payments, the architecture decisions above are best made before build, not during assessment. Our enterprise integration services cover API and event platform integration, legacy modernisation and migrations, delivered by senior nearshore teams with experience in regulated, high-volume environments. For the AI side, see our four decisions that make AI system integration enterprise ready.

To discuss scope reduction, segmentation or a compliant integration design for your estate, start from Polarising.
FAQ
What are the requirements for PCI compliance?
PCI DSS v4.0.1 has twelve principal requirements covering network security, secure configuration, protection of stored and transmitted account data, malware protection, secure development, access control, authentication, physical security, logging, regular testing and security policy. Which ones you must validate, and how, depends on your role and the programme set by your acquirer or card brand.
Who is responsible for PCI DSS compliance?
The organisation that stores, processes or transmits card data, or can affect the security of that environment, is responsible. Service providers share responsibility for the parts they handle, and Requirements 12.8 and 12.9 govern how you manage them and how they evidence their controls. Using a provider never transfers your accountability.
Does moving to the cloud make us PCI compliant?
No. Cloud providers secure their platform and can supply compliance reports and a responsibility matrix, but you remain responsible for your configuration, access, integrations and applications. Map each requirement to provider, customer or both for every service you use.
Can we use AI in a payment environment?
Yes, if it meets the same requirements as any other system component. Keep card data away from models through tokenisation or masking, keep a human accountable for the process, and log AI actions under Requirement 10.
Sources
- PCI Security Standards Council: PCI DSS
- AWS Security Reference Architecture: PCI DSS
- Payment Card Industry Data Security Standard v4.0.1
- SecurityMetrics: A Guide to New Requirements in PCI DSS 4.0.1
- Checkout.com: PCI DSS 4.0.1, what merchants need to know
- PCI SSC: Scoping and Segmentation Guidance for Modern Network Architectures
- PCI SSC: Scoping and Segmentation announcement, December 2016
- Energize Global: Payment Security Architecture
- PCI SSC Cloud Computing Guidelines v3
- Migrating PCI Workloads to VMware Cloud on AWS
- PCI Watch: Understanding PCI DSS Third-Party Service Providers
- PCI SSC: AI Principles for payment environments
- PCI SSC: Additional guidance on securing AI in payment environments
Recommended
- 3 Moves Enterprise Engineers Must Take to Keep AI PCI DSS Compliant
- Enterprise Integration Services
- Banking Systems Integration: Architecture, Risks and Approach
- Vector Database Integration: An Enterprise Architecture Guide
Content strategy and optimisation for Google and AI-powered search platforms by Velveteen SaaS — specialists in online visibility and customer acquisition.
