AI Glossary
Data privacy in AI
data protection in AI, AI data confidentiality
Data privacy in AI is the set of rules and measures protecting personal and confidential data at every stage of working with a model: in training, in queries, and in responses. It defines what may be passed to the model and how long it is retained.
- Concerns data in training, in queries to the model, and in its responses.
- Key questions: what reaches the model, whether it is used for further training, and how long it is retained.
- Measures include data minimization, anonymization, access control, and clear retention rules.
Data privacy in AI concerns the protection of personal and confidential data throughout all interaction with a model. The risk appears in three places: in the training data, in the content of queries sent to the model, and in its responses, which may reveal information they should not contain.
In practice you have to answer concrete questions: what data reaches the model at all, whether the provider uses it for further training, and how long it is retained. The measures applied include data minimization, anonymization, access control, and retention rules. These arrangements are part of AI governance and are checked during an audit.
Related terms
In guides