Google DeepMind

Lyria 3

• 2 safeguards
• 2 source documents

Safeguards the provider describes

• Output provenance and watermarking

  product-level mitigations such as safety filtering and applying SynthID watermarking.

  Google DeepMind · 18 February 2026 · Read the source

• Mitigations include, but are not limited to : dataset filtering; conditional pre-training; supervised fine-tuning; reinforcement learning from human and critic feedback; safety policies and desiderata; product-level mitigations such as safety filtering and applying SynthID watermarking.

  Google DeepMind · 18 February 2026 · Read the source

No document we read states: a release date, a safety framework, whether a capability level was crossed, an independent or government test, a test by the provider. No incident on record names this version.

Compare with other versions