Google Develops Watermarking System for AI-Generated Proteins
Researchers at Google have developed a method to watermark AI-generated protein sequences, addressing growing biosecurity concerns as AI tools become increasingly capable of designing novel proteins.
The technique works with popular AI protein design tools, embedding invisible signatures into the amino acid sequences that can later be detected without destroying or altering the protein. This allows verification that a given protein was computationally designed rather than naturally evolved.
As AI-powered protein design accelerates—enabling applications from therapeutics to industrial enzymes—the ability to trace and authenticate designed proteins becomes critical for biosecurity oversight. The watermarks are engineered to survive common protein expression and purification processes while remaining detectable through straightforward analysis.
The approach represents a proactive step toward responsible AI development in synthetic biology, giving researchers and regulators a potential tool to distinguish AI-designed sequences from those found in nature.