Models learn from people. Every training set contains traces of real individuals, and every prediction touches someone whose data shaped the system. Data privacy addresses how that information is collected, stored, used, and protected.

Regulations set the floor. GDPR in Europe, CCPA in California, and similar laws elsewhere grant rights to access, correct, and delete personal data. AI systems complicate compliance because models absorb patterns from data and cannot easily forget specific examples.

Key concerns

• Consent and transparency in data collection
• Re-identification from supposedly anonymous data
• Model memorization of training examples
• Cross-border data transfer
• Inference of sensitive attributes

Techniques like differential privacy, federated learning, and data minimization reduce risk but add cost and complexity. Perfect privacy is rarely achievable. The practical goal is meaningful protection with clear accountability.