Google DeepMind adds hidden AI watermarks to proteins for traceability
Google DeepMind researchers in London have developed SynthIDBio, a system that embeds a detectable marker in AI-designed proteins’ amino-acid sequences and 3D shapes. Nature reports the watermark could help DNA-ordering companies and researchers identify synthetic proteins for database integrity and biosecurity, without noticeably reducing protein function.
Bottom line — SynthIDBio offers DNA-screening firms a new clue, but Nature reports that redesigning a protein can erase its watermark.
Go deeper 5
-
Nature reports that the watermark can be detected only with a secret key, intended to be shared with trusted partners such as DNA sequence makers.
-
In tests described by Nature, watermarked proteins bound to targets involved in viral infection, blood-vessel formation and immune regulation as efficiently as unwatermarked versions.
-
Chemistry World says DeepMind built separate models to watermark protein sequences and structures; the researchers found little change to protein interactions and structural geometry.
-
Nature cautions that running a watermarked protein through another design tool can produce a new sequence with the same structure and function but an obscured origin.
-
Biosecurity researchers quoted by Nature frame watermarking as one part of a broader defence, not a standalone safeguard.