Google DeepMind created a way to watermark AI-generated proteins
The new method adds digital signals to synthetic proteins to identify them as machine-made while keeping their biological function intact.
Google DeepMind released details on a new method called SynthID Bio to track proteins created by artificial intelligence. Researchers designed this system to embed digital watermarks directly into the structure of synthetic proteins.
Proteins are the building blocks of life, and AI is increasingly used to design new ones for medical research. Scientists need a way to distinguish these lab-made molecules from those occurring naturally in biology.
SynthID Bio serves as a proof of concept for watermarking AI-generated proteins while preserving biological function.
SynthID Bio works by altering the protein sequence in a way that remains invisible to the final shape and function of the molecule. The system acts as a verification tag that allows researchers to scan and confirm if an AI model generated a specific protein sequence.
Testing showed that these markers survive common laboratory processes. The team confirmed that the internal logic of the protein remains stable despite the hidden digital signature added during the generation process.
How do you think tracking the origin of artificial molecules will change the future of drug discovery?
This technology is currently a proof of concept. The source does not specify how it performs against advanced attempts to strip or bypass the watermarking signal in complex protein databases.
Source: Google DeepMind Blog