AI agents: ANSSI guidance also matters for GDPR compliance
In brief
On 30 October 2026, ANSSI published a two-page document titled Sécuriser un système IA agentique in its Essentiels collection. It is aimed at architects, CISOs, operators and developers and sets out seven principles. According to Journal du Net, AI agents can query databases, call business software, send messages or change records, often without a human at every step.
Source : Journal du Net - « Agents IA : pourquoi le dernier document de l'ANSSI concerne aussi les DPO »
What it changes for an SME
If you deploy agents that act on data, this note is not a new statute, but it already overlaps with the GDPR: minimisation, security of processing and data protection by design. According to ANSSI, you should analyse risks from design, update the analysis at each change, name a human owner, and hold back go-live while a risk remains unacceptable. You should also document each agent's mission, limit its access to a real business need, and class every action as autonomous, human-approved through a mechanism the agent cannot bypass, or forbidden. Security requirements from that analysis should appear in supplier specifications, beyond a GDPR Article 28 processor agreement.
What you can do
- List AI tools that can read, change or delete data without human approval at each step.
- Document each agent's mission and access rights, then class its actions as autonomous, approved or forbidden.
- Align this risk analysis with your DPIA and put the security requirements into supplier specifications.
Let's talk about the time you could save.
In 20 minutes, we find where AI could save you time or money. Free, no commitment.