Google DeepMind has developed a way to embed watermarks in proteins designed by artificial intelligence, according to Ars Technica. The technique is being called SynthID Bio, per Google DeepMind. Ars Technica reports that the method gives AI-designed proteins a verifiable marker. The work was reported on Wednesday, September 30, 2026. Ars Technica describes it as a method for watermarking AI-designed proteins. Google DeepMind's own material frames SynthID Bio as watermarking methods for synthetic biology. Unite.AI reports that DeepMind embeds verifiable watermarks in AI-designed proteins. The Nuclear Threat Initiative also weighed in, publishing a piece titled "Innovation Enables Responsibility: Watermarking to Strengthen DNA Synthesis Screening." That framing places the technique in the context of screening DNA synthesis, the process by which synthesized genetic material is ordered and produced. Taken together, the reporting points to a specific use case: marking AI-generated protein designs so they can be identified. Ars Technica's headline states that Google has figured out how to do this. The Nuclear Threat Initiative's headline links watermarking to strengthening DNA synthesis screening, suggesting the marker could play a role in that screening process. The material available does not detail how the watermark is embedded, how it is detected, or whether it has been tested outside Google. Those specifics are not included in the reporting summarized here. What to watch next: whether SynthID Bio is adopted by DNA synthesis providers and how detection works in practice.