{
  "id": 17416,
  "title": "Our approach to bioresilience",
  "url": "https://urgent.news/2026/07/16/our-approach-to-bioresilience",
  "topic": "ai",
  "section": "AI",
  "published": "2026-07-16T09:30:42.000Z",
  "source": {
    "name": "Google DeepMind",
    "slug": "google-deepmind",
    "url": "https://deepmind.google/blog/our-approach-to-bioresilience/"
  },
  "original_language": "en",
  "account": "In the rapidly changing global biosecurity landscape, the convergence of shifting natural ecosystems, widespread travel, and the potential misuse of AI necessitates heightened vigilance. However, AI serves as a crucial tool in our response efforts. Google DeepMind and Isomorphic Labs have unveiled their joint approach to bioresilience, aimed at preventing malicious use of their models, as well as empowering governments, scientists, biosecurity experts, and their own teams to harness these technologies for a more resilient world.\n\nOver the past year, the companies have forged more than 15 partnerships with governmental bodies, biosecurity organizations, and research groups to counteract threats, detect outbreaks promptly, and respond efficiently. They firmly believe that society must leverage AI's growing capabilities to tackle infectious diseases and prepare for future outbreaks. Previously inconceivable breakthroughs, such as DeepMind's AlphaFold, Isomorphic Labs' AI-powered Drug Design Engine (IsoDDE), and AlphaGenome, are revolutionizing the understanding of proteins, drug design, and genome function, enabling more proactive defenses against potential risks.\n\nTo ensure their models are safe and useful for experts, Google DeepMind employs a four-step safety process: threat modeling, evaluations, mitigations, and monitoring. They collaborate closely with in-house biologists, security experts, and external partners to comprehend potential threats, test their models against them, and implement robust safeguards. Additionally, they are adapting their SynthID watermarking technology, which could assist DNA synthesis providers in identifying risky AI-generated biological sequences.\n\nGoogle DeepMind's agent AlphaEvolve aims to optimize algorithms for producing and analyzing metagenomic sequencing data, enabling faster and more accurate DNA analysis. This cost-effective approach enables large-scale disease tracking. They are likewise investigating the application of AlphaGenome and Protein Function annotation technologies to identify novel patterns and emerging threats more swiftly than traditional methods.\n\nBy granting trusted researchers access to Google DeepMind's latest AI systems, Isomorphic Labs intends to accelerate the design of vaccines and other countermeasures against known and novel threats. Isomorphic Labs has also established a dedicated unit to rapidly deploy its drug design engine, IsoDDE, to develop medical countermeasures in response to both naturally occurring pandemics and potential risks arising from the misuse of advanced AI. The organization collaborates with governments and global health authorities to advance numerous diagnostic and therapeutic strategies, ultimately enhancing the real-world impact of the drug design engine for bioresilience.\n\nThis complex, long-term endeavor is part of their broader approach to managing Chemical, Biological, Radiological, and Nuclear (CBRN) risks, following the proactive mitigations and rigorous evaluation protocols of the Frontier Safety Framework. Google DeepMind and Isomorphic Labs are committed to open and collaborative work with biosecurity labs, governments, and the scientific community to ensure AI's responsible development and deployment to protect against one of society's greatest risks. More information on their efforts can be found in the full update on their approach.",
  "summary": "Google DeepMind and Isomorphic Labs are sharing our joint approach to bioresilience and AI models.",
  "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."
}