Google Explores SynthID Watermarks for AI-Generated DNA

Artificial intelligence is accelerating research into protein structures, genome function and drug discovery through systems including AlphaFold, AlphaGenome and Isomorphic Labs’ Drug Design Engine. The same capabilities have intensified concerns that advanced models could be misused to design hazardous biological material. Google DeepMind and Isomorphic Labs are therefore treating bioresilience as a core safety priority, seeking to reinforce existing screening by DNA synthesis providers while preserving legitimate scientific access.

The companies unveiled their joint bioresilience approach on July 16, 2026, saying they are working to adapt SynthID watermarking for biology so synthesis providers can identify potentially risky, AI-generated sequences. The program combines a four-stage safety process for models such as Gemini — threat modeling, evaluations, mitigations and monitoring — with tools including AlphaGenome for sequence analysis. It spans prevention, detection and response, and follows more than 15 partnerships advanced with governments, biosecurity organizations and research groups over the past 12 months.