What are the seven Privacy by Design principles?

They are proactive not reactive, privacy as the default setting, privacy embedded into design, full functionality (positive-sum not zero-sum), end-to-end security, visibility and transparency, and respect for user privacy (user-centricity). The framework, developed by Ann Cavoukian, underpins the GDPR's data-protection-by-design requirement.

How is Privacy by Design different for AI?

AI strains every principle — models can memorise training data, inference is hard to make transparent, and large data appetites pull against minimisation. The AI-specific items here cover things like training-data minimisation, avoiding model memorisation of personal data, explainability of automated decisions, and giving users control over how their data trains the system.

Does a high score mean I am GDPR compliant?

No. This checklist is a structured design aid, not a compliance certification or legal advice. Data-protection-by-design is one requirement among many, and compliance depends on your full processing context. Use the output to inform a proper DPIA and review with a privacy professional.

When should I use this?

As early as possible — Privacy by Design is about building privacy in before you write code, not bolting it on afterwards. Revisit it whenever the architecture changes materially, for example when you add a new data source or change how the model is trained.

Is my progress saved?

Yes — ticks and your system note are stored locally in your browser so a refresh does not lose them, and nothing is sent to a server.

AI Privacy by Design Checklist

AI privacy by design

Privacy by Design is the principle — now embedded in the GDPR as “data protection by design and by default” — that privacy should be built into a system’s architecture from the start rather than patched on later. AI makes this harder: models can memorise personal data from training sets, inference is opaque, and the appetite for data pulls directly against minimisation. This checklist evaluates your AI system against all seven Privacy by Design principles, each with concrete, AI-specific implementation items, and scores how well you cover each one.

How it works

You jot down your system architecture to keep the assessment concrete, then work through seven sections — one per principle. Under “privacy as default” you confirm the least-privacy-invasive settings are the defaults and training opt-out is on by default; under “embedded” you confirm data minimisation in training and inference; under “end-to-end security” you confirm encryption and protection against model-inversion and membership-inference attacks; and so on. Each section shows its own completion score so you can see which principle is weakest, and you can export the full scored assessment into your data-protection-by-design record or DPIA.

Tips and notes

Start before you build. Privacy by Design is a pre-code discipline; the cheapest privacy fix is the one you never had to retrofit.
Minimise training data, not just inputs. A model that memorises personal data is a privacy liability long after the data was “deleted.”
Make automated decisions explainable. Visibility and user-centricity both demand that people can understand and contest AI decisions about them.
Default to opt-out of training. Privacy-as-default means a user should not have to act to keep their data out of your model.