AI Privacy encompasses the principles, practices, and technologies used to protect personal and sensitive information as it flows through AI systems. It addresses privacy concerns at every stage: data collection, model training, inference (runtime usage), and output generation.
Key AI privacy challenges include: training data privacy (models may memorize and regurgitate personal information from training data), inference privacy (user prompts may contain sensitive personal data), output privacy (AI-generated content may inadvertently reveal private information), cross-context privacy (information shared in one context appearing in another), and metadata privacy (usage patterns revealing sensitive information about individuals or organizations).
Privacy-enhancing technologies for AI include: differential privacy (adding noise to training data to prevent individual identification), federated learning (training models on distributed data without centralizing it), homomorphic encryption (processing encrypted data without decryption), data anonymization and pseudonymization, secure multi-party computation, and on-device AI processing that keeps data local.
Regulatory requirements impacting AI privacy include GDPR's right to explanation for automated decisions, data minimization principles, purpose limitation, storage limitation, and the right to erasure — all of which must be addressed in AI system design and deployment.
