Privacy risk, assessed · a service line of the AI Security & Cyber Risk practice readiness & advisory  ·  senior-only bench
§ AI Security & Cyber Risk·Privacy / DPIA for AI·v2026.07

Know whether your AI system needs a DPIA, and produce one that holds up.

Privacy law caught up to AI. A data protection impact assessment (DPIA) under GDPR Article 35, a privacy risk assessment under the CCPA and CPPA automated decisionmaking technology (ADMT) rules, an impact assessment under the Colorado AI Act, and the profiling and automated-decision provisions spreading across the US state privacy laws all now reach the systems your teams are building or buying. The Privacy / DPIA for AI Systems assessment is a fixed-scope engagement that determines which regimes are triggered for a given AI system, adapts a proven DPIA and PIA methodology to AI-specific risk, and hands your privacy office a defensible, ready-to-finalize assessment and scope outline. It is readiness work you own, not legal advice.

Part of the AI Security & Cyber Risk practice · fees fixed in writing after scoping. See engagement models.

§ What this is

This is an assessment, not legal advice. We adapt the established GDPR Article 35 DPIA and privacy impact assessment methodology to the specific risk factors of an AI system, map it against the frameworks that apply to you, and produce a structured, ready-to-finalize assessment. DSE does not provide legal advice, does not certify compliance with the GDPR, the CCPA, the Colorado AI Act, the EU AI Act, or any other law, and does not guarantee any regulatory, supervisory, or enforcement outcome. Legal conclusions, including whether a specific processing activity is lawful or an assessment is legally sufficient, are for your counsel and your data protection officer. The assessment and the scope outline are yours to own; we work alongside your privacy, legal, and security teams.

§A
The analysis.
The risk factors a DPIA for AI must cover.

A DPIA and PIA methodology adapted to how AI systems actually create risk.

A general DPIA template does not surface the failure modes of a model. We work each dimension a privacy assessment for an AI system has to address, mapping GDPR DPIA requirements and US state risk-assessment structure to the AI-specific factors that drive the risk.

Trigger analysis

Which frameworks the system actually triggers

A jurisdiction-by-jurisdiction read of where the system operates and whose data it affects: whether GDPR Article 35 requires a DPIA, whether the CCPA and CPPA ADMT rules require a risk assessment, whether the Colorado AI Act treats it as high-risk, and how the profiling and automated-decision provisions across the US state privacy laws apply. We tell you what is triggered and why.

Automated decisions

Profiling and decisions with real effects

Where the system profiles people or informs decisions that carry legal or similarly significant effects, we frame it against GDPR Article 22 and the ADMT and consequential-decision concepts in the US state rules. This is usually what moves a system from routine to high-risk, and it drives both whether an assessment is required and how deep it must go.

Sensitive data

Special-category and sensitive personal data

We map the personal data the system ingests, infers, and outputs, with particular attention to special-category data under GDPR Article 9 and sensitive personal information under the US state laws. Large-scale or inferred sensitive data is an independent DPIA trigger and a driver of residual risk that a model-agnostic template routinely misses.

Necessity & proportionality

Necessity, proportionality, and data minimization

The core of a DPIA is whether the processing is necessary and proportionate to a legitimate purpose. For an AI system that means examining training and inference data, retention, purpose limitation, and whether the same outcome is reachable with less or less sensitive data. We document the analysis so your DPO can stand behind the conclusion.

Rights & freedoms

Risks to rights, and fundamental-rights framing

We assess the risks to the rights and freedoms of data subjects the AI system creates: discrimination and unfair outcomes, opacity, loss of control over data, and error at scale. Where a fundamental rights impact assessment (FRIA) framing under EU AI Act Article 27 is relevant to a deployer, we structure the analysis so it can feed that obligation.

Mitigations

Safeguards, mitigations, and residual risk

For each risk we identify the safeguards and mitigations that reduce it, human-review gates, access and logging controls, transparency and notice, minimization, and testing, and we state the residual risk that remains. That residual-risk statement is what tells your privacy office whether the system is ready and whether prior consultation with a supervisory authority is a question for counsel.

§B
The boundary.
Assessment, not a legal opinion.

A defensible assessment, with a hard line on what it is not.

The value is a rigorous, well-documented assessment your privacy office and counsel can finalize and rely on. Being precise about where readiness work ends and legal judgment begins is the whole point.

What the assessment is

A structured DPIA / PIA for an AI system

  • A fixed-scope DPIA and privacy impact assessment for a defined AI system, adapting the GDPR Article 35 methodology and US state risk-assessment structure to AI-specific risk factors.
  • A jurisdiction-by-jurisdiction trigger analysis across the GDPR, the CCPA and CPPA ADMT rules, the Colorado AI Act, and the profiling and automated-decision provisions in the US state privacy laws.
  • A documented risk assessment, mitigation set, residual-risk statement, and scope outline, all readiness artifacts you own and finalize with your DPO and counsel.
What the assessment is not

Not legal advice, not a certification

  • Not legal advice and not a legal opinion. Whether processing is lawful, or an assessment is legally sufficient, is a conclusion for your counsel and DPO, not for us.
  • Not a certification of compliance with the GDPR, the CCPA, the Colorado AI Act, the EU AI Act, or any other law. We are not an accredited certification body.
  • Not a guarantee of any regulatory, supervisory, or enforcement outcome, and not a substitute for prior consultation with a supervisory authority where your counsel determines one is required.
§C
The format.
Scoped, senior-led, per system.

How the assessment runs.

A short scoping agreement up front, a focused mapping-and-analysis window against the frameworks that apply, and a documented assessment your privacy office can finalize. Built for a system going into or under review, not a multi-quarter program.

§ The engagement·scope · map · assess · document

We keep it tight and senior-led, scoped to the system and the jurisdictions in question and agreed in writing before analysis begins:

Want a fast read before you scope? Start with the free DPIA Threshold Checker. It runs entirely in your browser.

§D
The deliverable.
A finished, ready-to-finalize assessment.

You leave with a DPIA you can finalize and defend.

The engagement ends in artifacts your DPO, your privacy counsel, and your risk committee can use: a clear read on whether an assessment is required, a rigorous assessment of the risk, and the residual-risk position the system rests on. Not a certificate. A defensible starting point you own.

§ You receive

A concrete, documented privacy read on the AI system the day the engagement closes:

  • A jurisdiction-by-jurisdiction trigger analysis: whether a DPIA, a PIA, an ADMT risk assessment, or an impact assessment is required under the GDPR, the CCPA and CPPA ADMT rules, the Colorado AI Act, and the other US state privacy laws that apply, each with the reasoning.
  • A completed DPIA / PIA for the system: a structured assessment following GDPR Article 35 and the US state risk-assessment format: processing description, necessity and proportionality, data-subject rights, and the risks to rights and freedoms.
  • A risk register and mitigation set: each identified privacy risk with its likelihood and severity framing, the safeguards that reduce it, and the owner and sequence to put them in place.
  • A residual-risk statement: the risk that remains after mitigation, framed so your DPO and counsel can decide readiness and whether prior consultation with a supervisory authority is a question to escalate.
  • A scope outline and a brief for the privacy office and risk committee: what was assessed, what it means, and what to finalize and decide, in language a non-specialist reader can act on.
§ Scoped options·portfolio program & fundamental-rights framing

A single-system assessment is the core engagement. Two scoped options extend it when the shape of the work calls for it:

Every option produces readiness artifacts you finalize and adopt with your DPO and counsel. None of it is legal advice, a certification, or a guarantee of any outcome.

§E
Who it's for.
The desk that answers for privacy.

Built for the office that owns privacy risk across the AI portfolio.

Sponsored by the privacy office, and scoped so it gives the data protection officer, the privacy counsel, and the compliance and risk leaders in regulated sectors a defensible assessment they can finalize, distinct from the technical security buyer and a natural partner to it.

Privacy office

DPO / Chief Privacy Officer

You answer for whether an AI system needed a DPIA and whether the one on file holds up. A trigger analysis and a completed, well-documented assessment turn "we think we are fine" into a defensible position you can put in front of a regulator or a board.

Legal

In-house privacy counsel

You carry the legal judgment on lawfulness and sufficiency. The assessment does the analytical groundwork, mapping frameworks, data flows, and risk, so your legal conclusions rest on a rigorous, AI-specific record rather than a generic template.

Risk & compliance

CRO / compliance in regulated sectors

You need AI systems assessed to a consistent standard before they ship or come under examination. A per-system assessment and a portfolio methodology give you comparable evidence and a residual-risk position for the systems your program answers for.

§ Also at the table·security & AI governance

Security and the CISO are natural partners: the safeguards a DPIA relies on, access controls, logging, minimization, and testing, are the same controls the security program owns, and the assessment gives both sides a shared record. And the AI governance owner needs a per-system privacy assessment to sit inside the wider governance program, so the DPIA, the model inventory, and the risk register all point at the same systems.

Start free · client-side

Start with the free DPIA Threshold Checker.

Before you scope an assessment, get a read on whether you even need one. Answer a few questions about where your AI system operates, what it does, and whether it touches sensitive data, and our free, 100% browser-local DPIA Threshold Checker returns a likely yes or no, the framework or frameworks that trigger it, and a scope outline to build from. Nothing you enter leaves your browser. Bring the output to a scoping call and we turn it into a completed assessment.

What does it cost? The fee is fixed in writing after a short scoping call. It scales with the complexity of the system and the number of jurisdictions and frameworks in scope, and with whether you assess a single system or a portfolio, not your headcount. See the non-binding market-estimate ranges for this and every DSE engagement on the pricing page.

See engagement models
§F
Questions.
Straight answers.

Common questions.

What privacy, legal, and compliance leaders ask before they scope a DPIA for an AI system.

What is a DPIA for an AI system?+

A data protection impact assessment (DPIA) is a structured analysis of the privacy risks a processing activity creates and the safeguards that reduce them. GDPR Article 35 requires one where processing is likely to result in a high risk to the rights and freedoms of individuals, which commonly includes systematic profiling with legal or similarly significant effects and large-scale processing of special-category data. A DPIA for an AI system applies that same methodology to the specific risk factors of a model: the data it trains and infers on, the automated decisions it drives, and the errors and unfairness it can produce at scale. DSE delivers it as a fixed-scope assessment. It is readiness work, not legal advice.

Do I need a DPIA or a PIA, and which laws trigger it?+

It depends on where the system operates, whose data it affects, and what it does. Under the GDPR, a DPIA is required where processing is likely to result in a high risk, and it is expected for systematic profiling with significant effects and large-scale special-category processing. Under the CCPA, the CPPA rules require a risk assessment for certain uses of automated decisionmaking technology (ADMT). The Colorado AI Act requires deployers of high-risk AI systems to complete an impact assessment. Many other US state privacy laws add profiling and automated-decision assessment duties. The assessment runs a jurisdiction-by-jurisdiction trigger analysis so you know what is required and why. Whether a given assessment is legally sufficient is a conclusion for your counsel.

Is this legal advice?+

No. This is a privacy readiness assessment. We adapt the GDPR Article 35 DPIA and privacy impact assessment methodology to an AI system and produce a documented, ready-to-finalize assessment. We do not provide legal advice, do not certify compliance with any law, and do not guarantee any regulatory, supervisory, or enforcement outcome. Legal conclusions, including whether processing is lawful, whether a DPIA is legally required in your specific case, and whether prior consultation with a supervisory authority is needed, are for your data protection officer and your counsel. We work alongside them.

How does this map GDPR Article 35 to AI-specific risks?+

Article 35 sets the structure of a DPIA: a description of the processing, an assessment of necessity and proportionality, an assessment of the risks to the rights and freedoms of individuals, and the measures that address those risks. We keep that structure and populate it with the factors that make an AI system different: training and inference data and its provenance, special-category or sensitive data under GDPR Article 9, automated decisions and profiling under GDPR Article 22, model error and unfair outcomes at scale, opacity, and the safeguards that mitigate them. Where a deployer FRIA under EU AI Act Article 27 is relevant, we structure the risk analysis so it can feed that assessment too. The framework is used as a reference and structure, not certified.

How much does an AI DPIA cost?+

The fee is fixed in writing after a short scoping call and scales with the complexity of the system, the number of jurisdictions and frameworks in scope, and whether you assess a single system or a portfolio, not your company size. We publish non-binding market-estimate ranges for every engagement, including this one, on the engagement models page.

Last reviewed: 2026-07-05 · Initial release. Privacy / DPIA for AI Systems is a service line of the AI Security & Cyber Risk practice. All work is assessment, readiness, and advisory: a DPIA and privacy impact assessment adapted to an AI system, a jurisdiction trigger analysis, a risk register, and a residual-risk statement. It is not legal advice, not a certification of compliance with the GDPR, the CCPA, the Colorado AI Act, the EU AI Act, or any other law, and not a guarantee of any regulatory or enforcement outcome. Statutory frameworks are referenced as structure, not certified. We work alongside your counsel and data protection officer.