{
  "id": 13957327,
  "title": "Why Google DeepMind is watermarking ‘Biology’",
  "url": "https://urgent.news/2026/10/12/why-google-deepmind-is-watermarking-biology",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-12T12:26:08.000Z",
  "source": {
    "name": "Times of India",
    "slug": "times-of-india",
    "url": "https://timesofindia.indiatimes.com/technology/tech-news/why-google-deepmind-is-watermarking-biology/articleshow/134892812.cms"
  },
  "original_language": "en",
  "account": "Google DeepMind, Google's AI research laboratory, has unveiled a new technology called SynthID Bio that embeds a hidden watermark into proteins generated through artificial intelligence. The primary purpose behind this innovation is to enhance biosecurity by enabling the tracking of protein design origins and mitigating the risk of misinformation in synthetic 3D structures within public databases. The watermark does not hinder the protein's functionality, ensuring its effectiveness remains unaffected.\n\nSarah Carter, a biosecurity policy expert and Principal at Science Policy Consulting, described the technology as \"an important piece of the puzzle for tracking the provenance of biological designs.\" By linking designs to the model developers, the watermarks enable developers to take charge of safety measures and allow synthesis providers to streamline screening processes for their customers who employ these models. SynthID Bio is currently offered as a research release, with open-source code and data accessible to the scientific community.\n\nWatermarking is a familiar concept from photos and documents, where a subtle mark indicates the creator. In the realm of AI, companies have introduced watermarking for AI-generated content, with Google's SynthID serving as an invisible marker for AI-created photos, videos, audio, and text. SynthID Bio extends this concept to biology by embedding a signature within AI-generated protein designs, including both written sequences and predicted 3D shapes. This invisible mark can be verified, working on both digital models and the physical proteins once synthesized.\n\nIn laboratory tests, DeepMind reported that the watermarked designs maintained their performance and natural variety compared to unwatermarked designs. Biosecurity, as defined by the scientific journal ScienceDirect, encompasses practices and protocols to prevent the transmission of harmful biological agents, infectious diseases, and invasive species. The COVID-19 pandemic serves as a stark reminder of the potential devastation caused by biological threats, emphasizing the need for improved biosafety and biosecurity measures.\n\nThe rapid advancements in AI are enabling scientists to tackle some of biology's most challenging problems, such as predicting protein structures (AlphaFold) and designing entirely new ones (AlphaProteo, ProteinMPNN). However, these tools also introduce new risks. AI can generate designs that may evade conventional checks when DNA is ordered from a laboratory, and mislabeled synthetic 3D structures in public databases can mislead subsequent research.\n\nTo address these concerns, DeepMind introduced SynthID Bio, which functions as a family of methods tailored to the specific type of data being marked. For protein sequences, SynthID Bio subtly influences amino acid selection, while for predicted 3D structures, it makes minute adjustments to atomic coordinates. These slight modifications create a distinct signal that can be detected later without compromising the protein's biological function, crucial for applications in medicine or research.\n\nDeepMind likens biosecurity to a \"Swiss cheese\" defense, where multiple layers of safeguards protect against potential vulnerabilities. SynthID Bio adds an additional layer to existing security measures, embedding the watermark within the biological design itself. This innovation helps screeners identify orders from trusted models with built-in safety features, automating the process of verifying whether an unfamiliar sequence originates from a known source or represents a potential engineered threat.",
  "summary": "Google DeepMind has launched SynthID Bio to bolster biosecurity measures by encoding watermarks into AI-generated proteins. This innovative approach enables users to trace the origins of synthetic designs while preserving their biological functions. It seeks to combat the spread of false information in public databases regarding protein structures and sequences, thus aiding researchers in…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}