# GDPR-compatible AI agents for business

> Design GDPR-compatible AI agents through purpose, legal basis, processors, EU hosting, minimisation, access controls, audit logs, and approvals.

For AI agents: curated site map at [https://www.devolute.org/llms.txt](https://www.devolute.org/llms.txt). Canonical HTML (preferred for search engines): [https://www.devolute.org/en/insights/open-source/gdpr-compatible-ai-agents](https://www.devolute.org/en/insights/open-source/gdpr-compatible-ai-agents).

- Language: en
- Published: 2026-11-15
- Markdown URL: https://www.devolute.org/en/insights/open-source/gdpr-compatible-ai-agents.md

## GDPR-compatible AI agents depend on the whole architecture

Models, data sources, tools, hosting, logs, and human approvals belong in one assessment.

An AI agent does more than process one prompt. It reads mail or documents, invokes models and APIs, stores state, and writes to target systems. GDPR compliance therefore cannot be assigned to one model or an EU server location.

## Workflow checklist

1. Define purpose, legal basis, data classes, and affected people.
2. Record controllers, processors, and subprocessors.
3. Review processing agreements, region, and international transfers.
4. Retrieve only required fields and pseudonymise before model calls.
5. Give the agent a distinct least-privilege identity.
6. Implement retention, deletion, and data-subject rights.
7. Log inputs, tool calls, approvals, and outcomes.
8. Assess whether a DPIA is required.

Consequential actions need human approval. Local inference can reduce external transfers but does not replace purpose limitation, access controls, or accountability.

Read [AI security](/en/insights/open-source/ai-it-security-business), [sovereign AI](/en/insights/open-source/sovereign-ai-small-business), [local OpenClaw](/en/insights/open-source/openclaw-local-mac-mini), and our [agent orchestration service](/en/agency/open-source-llm-agents-orchestration).
