21.3 Privacy

Privacy in AI is about protecting people's personal data throughout the system's life. Because AI models learn from large amounts of data, they can expose, memorize, or misuse sensitive information if it is not handled carefully. Key ideas include consent, data minimization (collect only what you need), and anonymization.

For example, an Ethiopian telehealth app that uses AI to triage symptoms holds very sensitive health records. Privacy means patients consent to how their data is used, the app stores only what is necessary, and names or phone numbers are removed or encrypted before any model training.