It can. If an AI agent causes a personal data breach that is likely to put people's rights and freedoms at risk, the controller must notify the supervisory authority without undue delay and, where feasible, within 72 hours of becoming aware of it. The agent being automated does not change that.
This is general information, not legal advice. Ask your privacy counsel about your situation.
What the text says
Article 33 of the GDPR says that in the case of a personal data breach, the controller shall notify the competent supervisory authority without undue delay and, where feasible, not later than 72 hours after having become aware, unless the breach is unlikely to result in a risk to people. If notification is late, it must come with reasons for the delay. A processor must tell the controller without undue delay after becoming aware.
The notification must at least describe the nature of the breach, the categories and approximate numbers of people and records, give contact details for the data protection officer or other contact point, describe the likely consequences, and describe the measures taken or proposed. Where all details are not available at once, information may be provided in phases. The controller must also document every breach, including facts, effects and remedial action.
Where agents complicate it
| Question | Why agents make it harder |
|---|---|
| When did you become aware? | The agent may have acted for days before anyone noticed |
| What was exposed? | Tool call logs may be thin or missing |
| Who is the controller? | The agent may call external tools run by other firms |
| How many people? | Prompts and outputs may hold data nobody mapped |
The awareness question is a legal judgment, so do not decide it in a chat thread. Record the time someone first suspected a problem, and bring in privacy early.
What to prepare before an incident
- Know which agents can reach personal data. Your agent registry should say.
- Keep tool call records with identity, action, target and time, so you can count affected records.
- Keep a draft breach template with the Article 33 fields listed above.
- Name the contact point who will appear in the notification.
- Agree who decides whether risk to people is likely.
- Check contracts with tool and model providers for how quickly they must tell you about their own incidents.
What to do in the first hours
- Stop the agent: disable the identity and tools.
- Preserve logs and configuration before changing anything.
- Write down timestamps for first suspicion, confirmation and containment.
- Estimate categories and approximate counts. The text allows approximations and phased updates, so do not wait for perfect numbers.
- Start the internal breach record whether or not you end up notifying.
Do not forget the people involved
Separate from the regulator, you may need to tell affected individuals when a breach is likely to result in a high risk to them. Article 34 covers that, and your counsel can advise on when it applies.
Where Safeguard fits
Safeguard offers compliance and evidence automation and agent identity, which help keep the records this process depends on. A free scan is at https://gold.safeguard.sh/scan.
Quick check
Does the clock start when the agent acted or when we noticed?
The text refers to when the controller became aware of the breach. Document your timeline and ask counsel how it applies.
Can we notify in stages?
The text says information may be provided in phases without undue further delay when it cannot all be given at once.
Sources
- GDPR Article 33: Notification of a personal data breach
- GDPR Article 34: Communication of a personal data breach to the data subject
Last reviewed: 19 October 2026.
