Skip to main content

COMPEL Glossary / catastrophic-forgetting

Catastrophic Forgetting

Catastrophic forgetting occurs when AI models trained on new data lose previously acquired knowledge.

What this means in practice

This is problematic for continuously learning systems that must retain foundational capabilities. For organizations, it can cause well-performing models to lose abilities after updates. In COMPEL, it is a key risk in Module 3.5, Article 11.