DPA Checklist for AI Infrastructure
25 verification questions for data processing agreements in enterprise AI agent deployments.
A Data Processing Agreement (DPA) for AI infrastructure must address ten areas that standard SaaS DPAs do not cover: prompt logging policies, environment separation (dev/staging/production), model provider chains, in-flight vs. at-rest data processing, RAG embedding data protection, third-country access to production data, professional privilege compliance, PII tokenization, Decision Layer audit trails, and verifiability of technical measures. This checklist translates the ten gaps into 25 concrete verification questions.
The detailed analysis of the ten gaps is available in the magazine article: Data Processing Agreements for AI Agents: What Your Standard Contract Misses.
A — Data Categories and Processing Purposes
1
Are prompt contents and model responses listed as distinct data categories in the DPA?
2
Is it established that content classification responsibility lies with the organization, not the provider?
3
Are embeddings/vectors classified as potentially personal data?
4
Does the DPA address special categories under Article 9 GDPR that may arise through user inputs?
B — Logging and Monitoring
5
Is request/response body logging disabled in the production environment?
6
What metadata is recorded (status codes, latencies, request IDs)?
7
Is debug logging verifiably disabled in production?
8
Are stack traces and error messages configured to exclude content data from logs?
9
Is verification of logging settings part of the release process?
C — Environment Separation and Access
10
Do separate environments exist (dev, staging, production) with distinct data policies?
11
Do dev/staging environments contain exclusively synthetic or anonymized data?
12
Is production access restricted to authorized roles within the EU/EEA?
13
Does a documented exception procedure exist for support cases involving data?
D — Model Providers and Sub-processors
14
Is the delineation clear: Which providers are sub-processors of the provider, and which operate in the organization's tenant?
15
Is content retention at model providers disabled?
16
Is the exclusion of training data usage contractually documented?
17
Where are the model endpoints located (EU region, US, other)?
E — Data Storage and Deletion
18
Is it established where persistent content data is stored (database, region, provider)?
19
What backup retention applies, and how are deleted data handled within backups?
20
Can individual users delete their own data within the application?
F — Regulated Industries
21
Does the DPA contain provisions for professional privilege compliance (§203 StGB or jurisdiction-specific equivalent)?
22
Are confidentiality commitments in place for all personnel with access?
23
Is PII tokenization available as an optional module?
G — Governance and Verifiability
24
Is an audit trail for agent decisions anchored as a contractual component?
25
Can technical and organizational measures be evidenced on request (configuration documentation, redacted log excerpts)?
This checklist is a requirements catalog from an architecture and governance perspective. It does not constitute legal advice. Legal review and formal assessment of the DPA is the responsibility of the controller's legal department or external counsel.
Frequently Asked Questions
Is this checklist a substitute for legal advice?
No. The checklist is a requirements catalog from an architecture and governance perspective. It helps ask the right questions of your AI vendor. Legal review remains the responsibility of your legal department.
What size organization is this checklist relevant for?
Any organization deploying AI agents with access to personal data — regardless of size. The questions about professional privilege and employee representatives are sector-specific and can be skipped if they do not apply.
Where do the 25 questions come from?
From practical experience with enterprise AI infrastructure projects. Each question addresses a gap that standard SaaS DPAs do not cover for AI infrastructure. The detailed analysis is in the magazine article.
How does your AI infrastructure score on the DPA check?
We assess your setup against the 25 requirements and identify where action is needed.
Schedule a conversation