Why Article 28 matters before an AI tool touches client data
If an AI provider processes personal data on your firm's behalf, check that the contract contains the required UK GDPR Article 28 terms. First map the provider's role for each feature and the data it receives; other roles and confidential material need their own assessment.
Processor terms record obligations; a signature does not prove that the provider follows them. Assess the role, contract, settings and operational evidence together. The ICO guidance checked for this draft is marked as under review following the Data (Use and Access) Act, so confirm the applicable position before relying on the checklist.
The short answer for professional services firms
Have the privacy reviewer establish the firm's and provider's roles for the actual prompts, files, outputs and features. Ask whether Article 28 processor terms are required and record the current source, binding terms and unresolved points before approving personal-data use. The vendor's label does not replace that assessment.
For law firms, accountancy practices, financial advisers, insurers, brokers and consultants, identifiable information should trigger a data-flow and role assessment, not an automatic processor designation. Record who determines the purposes and means for each activity. Different features or uses may need different assessments and contract provisions.
Check your AI vendor risk position
What a good AI DPA should cover
For processor arrangements, describe the subject matter, duration, nature and purpose of processing, personal-data types, data-subject categories and the controller's rights and obligations. Map prompts, uploaded documents, generated outputs, logs and support records separately so that the description matches the planned use.
At minimum, ask whether the provider will:
- process personal data on documented controller instructions, including any applicable UK-law exception;
- bind personnel who process personal data to confidentiality duties and agreed access limits;
- apply appropriate technical and organisational measures, assessing encryption, access control and monitoring against the processing risk;
- obtain prior specific or general written authorisation for sub-processors, with notice and an opportunity to object to intended changes under general authorisation;
- provide the required assistance with data-subject rights, security, breach notification, DPIAs and relevant ICO consultation, scoped to the processing;
- at the controller's choice, delete or return personal data at contract end and delete copies unless UK law requires storage; and
- ask which compliance information, audit and inspection rights apply to the assessed relationship; review the actual assurance documents against the intended product tier and record any gap.
The DPA should also be consistent with the product settings. If the contract promises no training on customer data, the admin console, retention settings and support process should line up with that promise.
The AI-specific questions to ask before signing
Check that the processing description covers the planned AI features and data types: prompts, attachments, embeddings, outputs, logs, user feedback, telemetry and support tickets. It should also ask whether any data is used to train, fine-tune, evaluate or improve models, including models run by third-party foundation model providers.
Key questions include:
- Are you a processor, controller or independent controller for each feature we plan to use?
- Which sub-processors and model providers receive our data, and in which countries?
- Can customer prompts, files or outputs be used for model training or product improvement?
- What retention periods apply to prompts, outputs, files, logs and backups?
- Can we disable features that expose client confidential or special category data?
- How do you evidence deletion, incident response and support access controls?
Give the data protection lead, risk owner and service owner enough evidence to assess the defined use. If an answer remains unclear, pause the affected activity. Any narrower pilot needs its own data and role assessment; reduced or synthetic data does not automatically resolve the missing evidence.
How this fits into an AI governance process
The DPA is one control inside a wider AI governance process. You still need a use case owner, a data classification decision, a DPIA or screening where relevant, access controls, user guidance, output review, evidence capture and a route for incidents or exceptions. The firm should be able to show who approved the tool, what data it may receive and what controls were checked.
Route regulated or client-confidential uses to the responsible reviewers. Assess data protection, security, confidentiality, privilege and applicable professional or sector duties separately, even if one intake form collects the evidence. A processor contract does not resolve privilege, Consumer Duty or professional-conduct questions.
Build a practical AI governance rhythm
Red flags that should pause the rollout
Pause the affected use where terms, sub-processor information, training rights, deletion, access controls or security evidence do not meet its requirements. Consumer or enterprise branding does not settle suitability. Public research prompts may still contain personal information or produce identifying outputs; assess the actual data rather than treating public availability as an exemption.
Another red flag is a mismatch between sales assurances and contract terms. If the salesperson says data is private but the contract allows broad product improvement use, treat the contract as the control point. Record the gap, ask for corrected terms and keep the tool out of sensitive workflows until the evidence supports the decision.
A simple approval checklist
Before approving an AI vendor for personal data, keep a short evidence pack:
- Use case description and data categories.
- Controller or processor role assessment.
- Binding Article 28 terms where the assessed relationship requires them; other arrangements reviewed separately.
- Sub-processor and international transfer review.
- Training, retention and deletion position.
- Security evidence mapped to the product tier.
- Named business owner and human review controls.
- Decision record with conditions, expiry date and review trigger.
This gives the firm an audit trail if a client, insurer, regulator or internal risk committee asks why the tool was allowed.
Conclusion
Before approving a use, reconcile the assessed provider role, binding terms where required, data-minimisation choices and product settings. Record evidence gaps and a responsible reviewer. Recheck the decision when the feature, processing purpose or contract changes; the DPA is not a finding of overall compliance.
Turn AI vendor checks into an implementation plan
Need a safer AI route for your firm?
If this article reflects a live decision in your practice, the useful next step is to map the workflows, confidentiality risks, supplier controls and governance gaps before tools spread informally.
