Google · Research · Safety and security
Google DeepMind unveils SynthID Bio, which adds invisible watermarks to AI-designed proteins
Google DeepMind unveiled SynthID Bio on September 30 alongside a paper in Nature. During protein sequence design, the method subtly steers amino acid choices to leave a mark that is invisible but statistically detectable; a fine-tuned AlphaFold 3 also embeds a mark in the 3D structures it predicts. In the lab, watermarked binders against VEGF-A, PD-L1 and SARS-CoV-2 worked as well as their unwatermarked counterparts. The aim is to help DNA synthesis providers and scientific databases recognize AI designs. The code is open on GitHub. Caveat: the team calls it a proof of concept and acknowledges that resequencing a design with another tool largely removes the mark.
Sources
- Nature, “Function-preserving watermarking of AI-generated proteins”, (nature.com)
- Google DeepMind, “Introducing SynthID Bio”, (deepmind.google)
- Google, “SynthID Bio watermarks AI-designed proteins”, (blog.google)
- GitHub (Google DeepMind), “google-deepmind/synthidbio”, SynthID Bio code repository (github.com)
- Nature (haber), “Secret watermark labels proteins as ‘made by AI’”, (nature.com)
About this story
This story was posted on Instagram by @jarrus.tech on Oct. 6, 2026.
Spotted an error in this story? [email protected] · Instagram
Short link: thejarrus.com/en/synthid-bio
This story in Turkish: Google DeepMind, yapay zekayla tasarlanan proteinlere görünmez filigran ekleyen SynthID Bio’yu tanıttı
On the same topic
Robot tests: even the best model completed only 19% of 84 robot tasks, and 63 tasks were solved by no model
In RobotWorld, a benchmark released October 7, the top model, GPT-6 Astra, completed only 16 of 84 simulated robot tasks; no model solved 63 of the tasks.
OpenAI’s image model posts a near-perfect score on text-dense images
On UltraText Bench, a dense-text test led by Westlake University, OpenAI’s GPT Image 2 ranked first of 24 configurations with 99.35 out of 100.
AWS adds Chinese lab Z.ai’s GLM 5.3 model to Amazon Bedrock
AWS added GLM 5.3, a 753-billion-parameter model from Beijing-based Z.ai (formerly Zhipu AI), to Amazon Bedrock for eligible enterprise customers on October 5.