AI systems require massive amounts of data — and that data often contains sensitive personal information. Privacy is one of the most acute ethical issues in AI deployment.

Key Privacy Risks

• Training data exposure: Models can memorize and regurgitate personal information from training data
• Inference attacks: Adversaries can extract sensitive information about training data from model outputs
• Surveillance at scale: AI enables mass surveillance of individuals previously impossible to achieve
• Data aggregation: Combining datasets reveals sensitive info not present in any single dataset

Mitigation Approaches

• Differential privacy, federated learning, data minimization

Reference:

EFF on AI and privacy

https://www.eff.org/issues/ai