Google DeepMind

Veo 3

• System card dated 23 May 2025
• 1 safeguard
• 1 test by the provider
• 1 source document

Safeguards the provider describes

• Output provenance and watermarking

  Post-training mitigations included applying tools such as SynthID watermarking and production filtering to reduce information integrity harms and minimise harmful outputs.

  Google DeepMind · 23 May 2025 · Read the source

The provider's own tests

• content safety testing · The provider · Harmful content

  In the course of testing for content safety, we identified areas of content policy violations and potential abuse, and specific issues were mitigated prior to launch.

  Google DeepMind · 23 May 2025 · 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. No incident on record names this version.

Compare with other versions