Regulated software compliance is not a certificate you collect at the end of a project. It is a set of controls, evidence and accountabilities built into how software is designed, integrated and changed, so that an auditor can see what happened and why. For enterprises with legacy estates, AI components and dozens of connected systems, the hardest part is rarely the rulebook. It is proving the rules were followed across every integration. That is the work our senior nearshore teams in the UK and Portugal take on, with governance and accountability designed in from the first sprint.
Key Takeaways
- Compliance means adhering to the laws, standards and internal policies that govern how software handles data, security and operations. Which ones apply depends on your sector, jurisdiction, data and contracts.
- Treat compliance as an engineering property: requirements in the backlog, controls in the pipeline, evidence generated automatically.
- Integration layers are where compliance most often breaks, because data crosses system boundaries that no single team owns.
- Rules are moving. The EU Cyber Resilience Act reporting duties began on 11 September 2026, and AI transparency duties under the EU AI Act took effect in August 2026.
- A checklist only works if each line has an owner, a piece of evidence and a review date.
Table of Contents
- What regulated software compliance actually means
- Which rules apply to your software
- Building compliance into the engineering lifecycle
- Why integrations are where compliance is won or lost
- A regulatory compliance checklist for software projects
- Making evidence a by-product of delivery
- How we approach regulated delivery
- Where to start
- FAQ
- Sources
What regulated software compliance actually means
Microsoft’s engineering playbook defines compliance as adherence to regulatory standards, legal requirements and organisational policies covering data handling, security practices and operational procedures. It names GDPR, HIPAA and PCI DSS as examples, alongside internal governance frameworks.
Two things follow for a board or an IT leadership team.
First, compliance has an external side (law, regulators, customer contracts) and an internal side (your own policies and risk appetite). A system can satisfy a regulator and still breach your own policy, and the reverse.
Second, it covers the whole lifecycle. A compliant application deployed through an uncontrolled release process, or fed by an undocumented integration, is not a compliant service.

Which rules apply to your software
Applicability is a function of industry, jurisdiction, the type of data you process and what your customers have written into contracts. There is no universal list, but a handful of regimes cover most enterprise estates.
| Regime | Where it applies | What it means for software |
|---|---|---|
| GDPR / UK GDPR | Personal data of people in the EU or UK | Privacy by design and by default, data minimisation, breach notification within 72 hours, impact assessments |
| PCI DSS | Anyone handling payment card data | Controls on systems that store, process or transmit card data |
| HIPAA | US health data held by covered entities and their suppliers | Administrative, physical and technical safeguards for electronic health information |
| SOX | US-listed companies | Change management, segregation of duties and audit trails on systems feeding financial statements |
| DORA | EU financial entities and their ICT suppliers | Documented ICT assets and functions, resilience and evidence |
| EU Cyber Resilience Act | Products with digital elements sold into the EU | Lifecycle security, vulnerability reporting, technical documentation |
| EU AI Act | AI systems placed on or used in the EU market | Obligations set by use case and risk level |
The GDPR details above, along with the HIPAA and SOX implications, are summarised in ISC2 CSSLP study material, which also gives fines of up to 4% of global annual turnover or EUR 20 million under GDPR.
The UK picture
If you operate in Great Britain, the questions are usually sector-led. Financial firms in scope of the FCA’s rules had until 31 March 2025 to show they could keep their important business services within impact tolerances, as the FCA’s operational resilience page sets out. That is a software and integration question as much as a policy one: you cannot map an important business service without knowing which systems and interfaces support it.
For critical national infrastructure and essential services, the NCSC’s Cyber Assessment Framework is outcome-based rather than a checklist of actions, and includes a specific objective on secure software development and support. Version 4.0 asks, among other things, how you manage supplier software.

Rules that moved recently
If you sell software or connected products into the EU, three dates matter:
- The Cyber Resilience Act entered into force on 10 December 2024. From 11 September 2026, manufacturers must report actively exploited vulnerabilities and severe incidents, including for products already on the market, and most other obligations apply from 11 December 2027, with fines of up to EUR 15 million or 2.5% of global turnover, according to MoFo’s European digital compliance briefing.
- The same briefing notes that AI Act transparency obligations took effect on 2 August 2026: systems that interact directly with people must say so unless it is obvious, and providers must mark synthetic content in a machine-readable way.
- The AI Act regulates uses rather than industries. According to Confir’s sector guide, stand-alone high-risk systems in areas such as employment, essential services and critical infrastructure face a 2 December 2027 deadline, while AI embedded in regulated products such as medical devices follows on 2 August 2028.
Customer-service-heavy sectors such as telecoms can meet the transparency duties well before any high-risk deadline, and a minimal-risk classification does not mean nothing is due.
In the US, California’s privacy regulator adopted updated rules in July 2025 that require certain businesses to carry out risk assessments and annual cybersecurity audits, and give consumers rights around automated decision-making technology. Those rules take effect on 1 January 2026, per the California Privacy Protection Agency.
Healthcare software
For health and medical device software, regulators expect specific cybersecurity deliverables. This overview of FDA cybersecurity deliverables for SaMD covers vulnerability management, lifecycle security and machine-readable software bills of materials for US-bound products. Even if you are not selling into the US, the same pattern of evidence is a useful model for any regulated health system.
Building compliance into the engineering lifecycle
The most useful common vocabulary is NIST’s Secure Software Development Framework. SP 800-218 organises practices into four groups: prepare the organisation, protect the software, produce well-secured software, and respond to vulnerabilities. It applies to any sector or organisation size, takes a risk-based approach to which practices to adopt, and prescribes outcomes rather than tools. That makes it a practical way for an acquirer to tell a supplier what it expects.
Policy, process and procedure
OWASP’s governance, risk and compliance capability separates three layers. Policies state expectations, processes are repeatable sequences, procedures are step-by-step instructions. Its authors warn that a policy that exists only as a PDF nobody reads changes no behaviour, and that strong policy with weak procedure produces inconsistent outcomes.
In practice, executives should ask for each regulatory obligation: where is the policy, which process implements it, and what procedure does an engineer follow on Tuesday morning?
Controls as code
A 2024 academic paper on continuous compliance automation describes embedding controls from code commit through deployment and runtime, mapping test results to control identifiers and enforcing in graduated steps. Translating a vague requirement into a testable assertion, such as “backups are encrypted”, makes it measurable and repeatable. Over time the tests become a living library that changes with your architecture and the regulation.
Why integrations are where compliance is won or lost
Most enterprise compliance failures are not in a single application. They sit between applications: a payment payload logged in plain text by a middleware step, a patient record replicated to an analytics store with no retention rule, an AI model called through an API nobody registered.
Integration layers concentrate risk for three reasons:
- Data crosses boundaries. The owner of the source system is not the owner of the consumer, so purpose limitation and retention rules get lost in transit.
- Change is constant. New APIs, events and mappings ship weekly, and each one can alter what data moves where.
- Evidence is scattered. Logs, approvals and test results live in different tools, so assembling an audit trail takes weeks.
That is why integration platforms matter to compliance. Whether your estate runs on Boomi, MuleSoft, TIBCO or Solace, the platform is where you can enforce consistent policy on every flow: authentication, schema validation, routing rules, audit logging. We work across these four platforms, and our integration platforms page describes that work. The design principle is the same on each: put the control in the shared layer, not in every individual application.
Governance of APIs needs the same discipline without slowing delivery, which we cover in our piece on API governance in legacy enterprise systems. For payment and AI workloads specifically, see our guidance on keeping AI systems PCI DSS compliant. For credentials and keys, see our article on infrastructure-first secrets management for UK teams.
A regulatory compliance checklist for software projects
A useful checklist is organised by project phase and every line carries a status, an owner, evidence and a deadline. Saudi Arabia’s national cybersecurity authority publishes an illustrative software development checklist built this way. Its planning phase includes software asset registration, a centralised development toolchain, threat modelling, evaluation of legal, regulatory and internal compliance, and application risk profiling. Its companion IT projects and change management checklist asks for the same at initiation and says the checklist should be reviewed at least annually or when technical, policy or regulatory changes occur.
Adapted for an enterprise integration programme, a working version looks like this:
Before you build
- Identify which regimes apply by sector, jurisdiction, data type and contract.
- Register the system, its interfaces and its data flows in one inventory.
- Classify the data and the risk of each application or AI use case.
- Run threat modelling and record a decision for every material risk.
- Name an accountable owner for each obligation.
While you build
- Separate development, test and production environments. Australia’s Information Security Manual makes this an explicit control (ISM-0400) and calls for secure-by-design principles supported by threat modelling.
- Enforce approvals and segregation of duties in the pipeline, not by email.
- Run static, dependency and dynamic security tests on every change.
- Generate a software bill of materials. The EU Cyber Resilience Act treats a machine-readable SBOM as part of a product’s technical documentation, according to SonarSource’s SBOM guide.
- Keep secrets out of code and configuration.
After you release
- Monitor integrations for drift: new consumers, changed schemas, unexpected volumes.
- Keep a tested process for reporting vulnerabilities and incidents within the deadlines your regimes set.
- Retain evidence for the period your regimes require. US public companies under SOX, for instance, must keep certain audit and financial records for at least seven years, per EBSCO Research Starters.
- Review the checklist at least once a year and whenever a rule changes.
Making evidence a by-product of delivery
Auditors ask for proof, not intentions. A well-run delivery pipeline already produces most of it: traceability of changes, approvals and policy gates, timestamped test results, signed artefacts and monitoring records. The challenge is keeping it tamper-resistant and retrievable. For DORA in particular, the regulated DevSecOps guide argues that pipelines generate this evidence naturally when designed for it.
For a CTO, the test is simple. Pick one production change from last month and ask the team to produce, within an hour, who approved it, what was tested, what was deployed and how you would roll it back. If that takes a week, your compliance is a documentation exercise rather than an engineering property.
How we approach regulated delivery
Our teams are senior, nearshore and based in the UK and Portugal, and our practices are certified and audited for regulated environments across finance, healthcare, telecoms, logistics, retail and technology. Our IMS policy sets out the management system behind that. We are strongest in legacy-heavy estates where generic teams struggle, because those are the estates where undocumented interfaces and inherited processes make compliance hardest.
The Polarising view
We believe compliance fails when it is owned by a committee and delivered by someone else. The people who design the integration should be the people who understand the obligation it carries, which is why we put senior engineers on regulated work and build governance and accountability into AI systems from the outset, not after go-live.
We also believe the integration layer is the most underused compliance control an enterprise has. Whether the estate runs on Boomi, MuleSoft, TIBCO or Solace, a consistent policy applied at the platform level beats a hundred separate fixes in individual applications. Our job is to connect the platform, the legacy systems and any AI components so that every flow can be explained to an auditor.
Our experience across finance, healthcare and telecommunications, including work such as the integration revamp for TAP Air Portugal and the transition for CTT Locky, taught us the same lesson each time: complexity is manageable when it is faced early and engineered openly.
Polarising Team
Where to start
If you are planning an integration, modernisation or AI programme in a regulated environment, our enterprise integration services cover API and event platform integration, legacy modernisation and migrations, delivered by senior teams with flexible engagement models that scale up or adjust seniority as the programme evolves. For AI components, our AI engineering services take work from pilot to production.

To discuss a regulated programme with our engineers, start at polarising.com.
FAQ
What is compliance in software engineering?
It is the practice of making sure software and the processes that build and run it meet the laws, standards and internal policies that apply to them. That includes how data is handled, how security is managed and how changes are approved and recorded. It is continuous, not a one-off sign-off.
What are the main types of software compliance?
Most enterprises deal with privacy law (GDPR), sector rules (HIPAA in US healthcare, PCI DSS for payments, SOX for financial reporting, DORA for EU finance), security standards such as ISO/IEC 27001 and NIST SP 800-218, and contractual obligations to customers. AI-specific rules, such as the EU AI Act, are now a further layer.
Does the NIST framework apply outside the US?
SP 800-218 states that it applies to any sector, organisation size, software type or platform, and uses a risk-based approach to choosing practices. Many organisations outside the US use it as a shared vocabulary for secure development requirements, even where it is not legally mandated.
How often should a compliance checklist be reviewed?
At least once a year, and whenever there is a significant technical, policy or regulatory change. That is the cadence the Saudi NCA’s project checklist sets, and it is a sensible default for any enterprise programme.
Sources
- Compliance, Engineering Fundamentals Playbook
- CSSLP: Regulations, Standards and Guidelines for Software
- FCA: Operational resilience
- NCSC Cyber Assessment Framework 4.0
- MoFo: European Digital Compliance
- EU AI Act by Industry: Sector Compliance Guides
- CCPA Updates, Cybersecurity Audits, Risk Assessments
- NIST SP 800-218, Secure Software Development Framework 1.1
- OWASP PSCF: Governance, Risk and Compliance
- Policy-as-Code, CI/CD Enforcement and ISCM
- NCA software development cybersecurity checklist
- NCA IT projects and change management checklist
- Australian Information Security Manual, software development guidelines
- SonarSource: SBOM for compliance
- EBSCO Research Starters: Software Regulations
- Continuous compliance in DevSecOps with CI/CD
Recommended
- API Governance Without Blocking Delivery in Enterprise Legacy Systems
- Resilience Engineering for the Enterprise: A Leader’s Guide
- AI Risk Management for Enterprise Leaders: Board Ready Registers
- 3 Moves Enterprise Engineers Must Take to Keep AI PCI DSS Compliant
Content strategy and optimisation for Google and AI-powered search platforms by Velveteen SaaS — specialists in online visibility and customer acquisition.
