1. What Article 32 of the GDPR requires
Article 32 of the GDPR (Regulation (EU) 2016/679) concerns the security of the processing of personal data. It obliges the controller and the processor to adopt technical and organisational measures appropriate to the risk. Among the examples the provision itself gives are pseudonymisation and encryption, the ability to ensure the confidentiality, integrity and availability of systems, the ability to restore data in a timely manner after an incident, and a regular process of testing the effectiveness of the measures.
The key word is 'appropriate'. The GDPR does not list a menu valid for everyone. It requires an assessment of the state of the art, the costs, the nature, scope, context and purposes of the processing, and the risks to individuals. For this reason two businesses of the same size may have different measures, and both can be compliant if the choice is justified and documented.
2. Start from risk, not from a list
The correct method starts with an inventory of the processing: which personal data you collect, from whom, where you keep it, who has access to it and for how long. For each processing activity, assess the possible harm in the event of loss, alteration, unauthorised access or unavailability. A customer database and an archive of event photographs do not have the same risk profile.
Only after this analysis does it make sense to choose the measures. Measures must be proportionate: a complex control is not needed for a register containing little non-sensitive data, whereas much more is required for health data or archives containing staff information. Document the assessment and the decision, because in the event of an inspection the controller must be able to demonstrate that the reasoning was carried out.
3. Encryption and pseudonymisation: what they really do
Encryption in transit protects data while it travels across networks, for example with TLS protocols on websites and email services. Encryption at rest protects data stored on disks, databases and backups. The two are different and must be verified separately. A website with an encrypted connection may still store data in clear text on an unprotected server.
Pseudonymisation replaces direct identifiers, such as a name or tax code, with codes that require a separate table to be linked back to the person. Pseudonymised data remains personal data, because it can be re-attributed. It should therefore not be treated as anonymous. The critical point is the management of the key or correspondence table: if it is kept alongside the data, the protection is only superficial.
4. Measures compared: what to verify
The table compares the most common technical measures with their purpose and with the practical check the business should be able to demonstrate. The entries are general: the concrete choice depends on the risk of each processing activity and on the existing architecture. Use the rows as a starting checklist and complete the verification column with your own company's data. Each row requires an owner and a date.
Every measure must be verified over time. Encryption that was active a year ago may have been switched off during a migration, and a shared administrative account may have remained active after a collaborator left. Periodic verification is exactly what Article 32 requires, so it should be documented as routine activity. If the verification leaves no written trace, for an inspector it has never taken place.
| Measure | What it is for | Practical check |
|---|---|---|
| Encryption in transit | Protects data while it travels across networks | Testing the configuration of sites and services |
| Encryption at rest | Protects disks, databases and backups | List of encrypted systems and key management |
| Pseudonymisation | Reduces direct identifiability in processing | Documented separation of the correspondence table |
| Access control and MFA | Limits who can view and change data | Periodic review of accounts and privileges |
| Backup and restore | Ensures availability after an incident | Restore test with the outcome recorded |
5. Restoration, backups and incident notification
Data availability is part of security. An incident such as ransomware, a failure or a human error can render data inaccessible even if nobody has read it. The 3-2-1 rule is a widely used starting point: three copies of the data, on two different media, one of them off-site. It is not a legal requirement, but it helps to think about restoration.
Restoration must be tested, not just planned. A backup that has never been restored is a hope, not a measure. In addition, in the event of a personal data breach, the GDPR requires notification to the supervisory authority within 72 hours of becoming aware of the breach, where the conditions set out in the regulation are met. Anyone handling data must have a procedure ready to meet this deadline.
6. Operational checklist for the controller
The list below is intended to take stock of the basic measures. It does not replace the risk assessment, but it indicates the questions to start from. Each item should have an owner and a verification date, otherwise it remains an intention. You can use it as the agenda for the first meeting with your supplier or your IT manager. There is no need to complete it in one day: what matters is having a timetable.
If several items remain unanswered, there is no need to buy new tools straight away. Often the first useful step is to define responsibilities, review access and test restoration. These three steps cost little and reveal a great deal. Start with the register of processing, then review access and finally test the restoration of a system. Each step produces a document that remains as evidence of the work done.
- Up-to-date register of processing, with data, purposes and places of storage.
- Encrypted connections for sites, services and email, verified with a testing tool.
- Encrypted disks on servers and laptops, with documented key management.
- Multi-factor authentication for remote access and for administrative accounts.
- Quarterly review of accounts, with removal of those no longer needed.
- Backups following the 3-2-1 rule, with one copy protected against accidental or malicious deletion.
- Documented restore test at least once a year.
- Written procedure for notifying a breach within 72 hours.
7. Common mistakes in technical compliance
The first mistake is buying a tool and considering the problem solved. Compliance cannot be bought in a box: an encryption tool is no use if the keys are stored in a file next to the data, and an antivirus does not replace access control. The second is encrypting without managing keys. Lost keys make data unusable, and poorly kept keys make encryption pointless.
The third mistake is treating pseudonymisation as anonymisation. Pseudonymised data remains personal data. The fourth is failing to review measures after a change: a new supplier, a new site or a new cloud service can undo protections that previously worked. For this reason verification should be repeated after every significant change, not only once a year. Ask your supplier which controls change when it modifies a service.
8. How to start: the first concrete step
The first step is the register of processing, even in a simple form. For each processing activity, write which data you process, who has access to it, where it is stored and for how long. This document is the starting point for the risk assessment and for all other choices. Even a simple version, updated every quarter, is much better than a perfect but outdated document.
Then choose the three processing activities with the most sensitive data and check the five measures in the table against each. Note what is missing, assign an owner and set a date. If your systems include an environment that has never been restored, schedule a test straight away. It is the check that most often reveals a real problem. Do not simply read about these measures: ask for proof.
Frequently Asked Questions
Does the GDPR impose a precise list of technical measures?
No. Article 32 requires measures appropriate to the risk and to the state of the art, and it cites some categories as examples. There is no list valid for everyone. Compliance depends on the risk assessment and its documentation. A business must be able to explain why it chose certain measures and how it verifies them.
Is encrypting data enough to be compliant?
No. Encryption is an important measure, but it must be accompanied by proper key management, access control, monitoring and backups. Data encrypted with a key kept alongside the data offers limited protection. Assess encryption within the overall context of your risks.
Does pseudonymisation exempt me from GDPR obligations?
No. Pseudonymised data is still personal data, because it can be linked back to a person with the right information. Pseudonymisation reduces risk and can be a useful measure, but the obligations remain. It should be assessed case by case by the controller.
Does my cloud provider guarantee compliance on my behalf?
Not entirely. The controller remains responsible for the choices made about its data, even when it entrusts the data to a provider. The provider can offer measures and documentation, but a contract is needed that defines roles, obligations and checks. Ask for evidence and do not accept generic statements alone.
How often should the effectiveness of measures be verified?
There is no universal interval, but verification must be periodic and carried out whenever an important system changes. For the most sensitive processing, at least an annual review is advisable, with documented restore tests. Record the checks and corrective actions: they are the evidence that the measures work over time.