Sensitive personal data includes health information and political opinions, which receive heightened protection under GDPR Article 9. Generative AI can also infer sensitive attributes from ordinary writing or accumulate behavioral patterns through persistent memory, creating profiles users may not expect. section 5.4.1 section 5.6.1
For autonomous agents, purpose limitation and data minimization are practical risk controls: an agent can drift from its original task into collecting or categorizing sensitive information. Disclaimers do not remove an organization’s accountability for personal-data accuracy; false information can also trigger rectification rights or affect significant decisions. section 5.4.1 section 5.6.1
Scheduling becomes health profiling
An agent asked to schedule a doctor’s appointment reads a discharge note mentioning diabetes, then creates an “endocrinology” profile tag for future use. The task has expanded from scheduling to inferring and categorizing sensitive health data. section 5.6.1
