Google’s DeepMind team has developed a SynthID-based method that embeds watermarks in AI-designed protein sequences without impairing their function. The system could help identify proteins created by trusted researchers and flag unmarked designs for closer biosecurity scrutiny.