SynthID Watermarking Technology Expands to AI-Generated Proteins

Google DeepMind expanded its SynthID watermarking technology to biological data, launching a proof of concept called SynthID Bio to track AI-generated proteins. The technology adapts the same core watermarking principles Google uses to identify AI-generated video in Google Veo 3 and synthetic images in Imagen 3, applying them to complex molecular structures.

What's new

SynthID Bio introduces a method for watermarking synthetic proteins by making minor, non-disruptive modifications to their amino acid sequences. Developed by Google DeepMind as of March 2025, this system embeds a digital signature directly into the protein's design. Crucially, the watermark does not alter the protein’s three-dimensional structure or its intended biological function.

The system works by utilizing the redundancy in the genetic code, where multiple codons can translate into the same amino acid. By strategically choosing specific codons during the AI generation process, SynthID Bio creates a verifiable pattern. Researchers can then scan a protein sequence to verify if it originated from a specific AI model, providing a critical layer of biosecurity and intellectual property tracking.

How it fits your workflow

While SynthID Bio directly benefits molecular biologists and biosecurity researchers rather than traditional filmmakers, its development signals a major shift in how Google manages AI output verification across all its models, including Google Veo 3. For creators concerned with digital provenance, the expansion of the SynthID ecosystem validates Google’s commitment to secure, traceable AI generation.

In the creative space, watermarking technologies like SynthID compete directly with open standards like the C2PA (Coalition for Content Provenance and Authenticity) used by Adobe and OpenAI. While C2PA relies on metadata and cryptographic wrapping, Google’s SynthID embeds the watermark directly into the content pixels, audio waves, or—in this case—genetic sequences. For creators using Google Veo 3 for AI video generation, the success of SynthID in high-stakes biological environments suggests that Google’s video watermarking will remain highly resilient to compression, editing, and format conversion.

What it costs / how to try it

Google DeepMind has released the research and code for SynthID Bio as an open-source project on GitHub. The tool is currently a proof of concept intended for academic researchers, biosecurity organizations, and developers working with protein-design models.