FRIA/DPIA · AI systems

FRIA and DPIA for AI systems

Bring privacy, fundamental rights, bias, human oversight and supplier information into one workable file. For AI systems that affect people, show high-risk signals or get stuck between GDPR, the AI Act and governance.

Impact sprint

2 weeks

Explain whether a DPIA, FRIA or combined file is needed

Show which rights and affected groups may be impacted

Challenge suppliers on missing information

Make a decision with visible residual risk

Where organizations get stuck

A DPIA focuses on privacy. A FRIA looks more broadly at fundamental rights. In AI projects those tracks often overlap, while legal, privacy, business and the supplier all hold different information.

1

Unclear obligation

Is this a DPIA, a FRIA, both, or first a classification decision? We make the sequence concrete.

2

Too little vendor evidence

Suppliers often provide generic security or model information, but not enough for bias, oversight and impact analysis.

3

No decision-ready file

Risks are spread across privacy, legal, procurement and business. We bring them into one reviewable decision memo.

What we deliver

DPIA/FRIA scope and decision on whether a combined file is useful

AI system context: purpose, workflow, users, affected people and supplier

AI Act role and risk route: provider, deployer, high-risk, transparency or low risk

GDPR link: personal data, special category data, legal basis, data flows and retention

Fundamental-rights analysis: discrimination, human dignity, access to services, explanation and appeal

Bias and fairness questions for data, model, use context and monitoring

Human oversight plan with authority, escalation, override and logging

Supplier questions and missing evidence per contract or product claim

Risk and mitigation matrix with owner, status and residual risk

Decision memo for legal, privacy, leadership or project steering group

Approach in 2 weeks

1

Scope and intake

We determine system, use case, team, affected group, AI Act route and the question: DPIA, FRIA or combined file.

2

Evidence request

We request focused documentation from internal teams and suppliers: data flows, model information, logging, oversight and instructions for use.

3

Impact workshop

Legal, privacy, business and process experts bring privacy and fundamental-rights impact together on one risk map.

4

Measures and vendor gaps

We translate risks into controls, human oversight, transparency, appeal route, monitoring and supplier questions.

5

Decision memo

We deliver a reviewable file with conclusion, open points, owners and go/no-go or follow-up advice.

Where this has the most impact

Public sector

For algorithms around public services, supervision, enforcement, benefits, education or citizen interaction.

HR and recruitment

For screening, matching, performance, planning, monitoring or other AI that affects workers or candidates.

Finance and essential services

For AI around credit, fraud, access to services, risk scores or customer decisions.

Legal, privacy and procurement

When a supplier does not sufficiently explain what the system does, which risks exist and who is responsible for what.

Afterwards you can

Explain whether a DPIA, FRIA or combined file is needed

Show which rights and affected groups may be impacted

Challenge suppliers on missing information

Make a decision with visible residual risk

Connect the outcome to the AI register, governance and evidence pack

Frequently asked questions

When is a FRIA needed?

A FRIA is especially relevant for high-risk AI systems and situations where fundamental rights may be affected. The exact obligation depends on the system, role, sector and use context.

What is the difference between a DPIA and a FRIA?

A DPIA focuses on privacy risks under the GDPR. A FRIA looks more broadly at fundamental rights such as non-discrimination, access to services, human dignity, explanation, appeal and effective human oversight.

Can DPIA and FRIA be combined?

Often yes. Especially for AI systems involving personal data, a combined file is more efficient, as long as privacy and broader fundamental rights are both treated explicitly.

What if the supplier provides little information?

Then we make that visible as a vendor gap. You receive concrete questions for contract, security, privacy, model governance, logging, bias and human oversight.

Is this legal advice?

This is practical AI governance and compliance support. For formal legal opinions, organizations work with their own lawyer or legal adviser where needed.

Make impact analysis decision-ready.

Start with the Gap Intake. We then determine which AI system, obligation and evidence belongs in the first FRIA/DPIA scope.

Rivium Westlaan 46, Capelle aan den IJsselCoC 90283597