Zum Hauptinhalt springen
Glossary

Fine-tuning

Updated on 1 min

Fine-tuning is the further training of an already pre-trained AI model on a smaller, organization-specific dataset. A general language or image model thereby learns the terminology, style or tasks of an organization without being trained from scratch. The data become part of the model weights and can hardly be removed afterwards.

This raises two questions. First, may the data be used for training? Personal data need a legal basis under the GDPR, and for customer data and protected works, contracts and usage rights need checking. Second, does fine-tuning change the role? Anyone who substantially modifies a high-risk AI system is treated as its provider under Art. 25 of the AI Act. The pattern MU-15 Further training with inadmissible data shows typical mistakes.

No Cookies!

This wiki collects nothing, bakes nothing and leaves nothing behind. There's nothing to consent to. Privacy doesn't get better than this.