Gemma Scope 2: helping the AI safety community deepen understanding of complex language model behavior

In three linesGoogle DeepMind releases Gemma Scope 2, open-source interpretability tools for the entire Gemma 3 family. These resources enable the AI safety community to deeply analyze language model behavior.

Summary generated by Claude — human-verified