desarrollo · 4 min read
Security in custom software: what your provider must guarantee
GDPR, backups, access control and auditing: what any custom software provider must put in writing before you sign the contract.
“What happens to our data if we switch providers, or if theirs suffers a breach?” is a question almost no SME asks before commissioning custom software, and one almost all of them wish they’d asked after an incident. At Evicron, an AI and custom software studio based in Barcelona founded in 2019 with over 200 projects delivered across 12 sectors, security isn’t a layer added at the end: it shapes, from the first sprint, who can see which piece of data and where that access gets logged.
Your provider is almost always a “data processor,” not just a technician
If the software you’re having built processes data about customers, employees or suppliers — and almost all of it does — the company commissioning it is the data controller, and whoever develops and maintains it acts as a data processor under the GDPR, as soon as they access that data during development or support. The European Data Protection Board’s (EDPB) SME data protection guide spells this out with a clear example: a payroll company acts as a processor when it only processes data on behalf of the SME, because it’s the SME that decides the purposes and means of the processing (EDPB). A custom software developer who only hands over code without touching real data might fall outside that scope; the moment they access a production database to debug an issue, they no longer do.
That distinction isn’t theoretical — it decides what the contract needs to say.
The processor agreement isn’t paperwork, it’s the minimum guarantee
Article 28 of the GDPR requires the relationship between controller and processor to be set out in a written contract, and Spain’s data protection authority (AEPD) publishes the minimum content such a contract must cover (AEPD): the subject and duration of the processing, the type of data and categories of people involved, the provider’s obligation to process data only on documented instructions, confidentiality for anyone with access, the security measures applied, the conditions for using sub-processors (any sub-processor needs prior authorisation and the same obligations), and what happens to the data once the contract ends — it must be deleted or returned, the company’s choice, unless a law requires otherwise.
If your provider’s proposal doesn’t mention any of this, it isn’t “already assumed” — it’s missing, and it’s negotiable.
What your provider must actually guarantee, not just promise on paper
Four areas are worth checking before signing, and worth asking about in numbers, not adjectives:
- Real backups. “We do backups” isn’t enough: ask for the exact frequency, where backups are stored, and above all the RTO (maximum time to restore service) and RPO (maximum data that could be lost), both in hours. A serious provider knows these figures by heart because they’ve tested them, not because they calculate them when you email asking.
- Role-based access control. Everyone involved — your team and the provider’s — should have access only to what their role needs, with an audit log showing who accessed which record and when. If anyone with a password can see the whole database, there’s no real access control, whatever a document promises.
- Encryption and environment separation. Production data shouldn’t be used unanonymised in a test environment, and it shouldn’t travel unencrypted between services. This is the most common cause of breaches that have nothing to do with a sophisticated attack: a copy of the real database left forgotten in an accessible development environment.
- What happens if there’s an incident. The GDPR requires notifying the supervisory authority of a breach within 72 hours of becoming aware of it. Your contract should fix how quickly the provider notifies you, because that clock runs whether or not the provider notices quickly.
Signs that security is just a word in the proposal
A few answers should raise a flag before you sign: “we’ve never had an incident” offered as the only reassurance (no one can promise that — only how they’d respond); not being able to say who, on their team, can access your production database; or treating security as a project phase instead of an ongoing practice — patching, dependency reviews, testing — throughout maintenance. None of these shows up alone in a bad hire: they pile up, the same way vendor lock-in does, which we cover in our guide on who owns the code in your custom software.
How we handle this at Evicron
In every custom software project we sign as data processor whenever it applies, with explicit Article 28 clauses rather than a generic template: what data we process, who on our team can access it, under what security measures, and what happens to that data if the project ends. Backups and test environments are separated from production from the first sprint, not added once an incident makes it necessary.
If you’re evaluating a custom software provider and want a checklist of what to ask for in the contract on security — or want a second opinion on one already on the table — you can get in touch and we’ll go through it in a first, no-cost conversation.