Google DeepMind and Isomorphic Labs have laid out a joint bioresilience program, formalizing how they intend to use artificial intelligence to prevent, detect, and respond to biological threats, including those that AI itself could enable.
The program, detailed in a July 16 blog post, is built on a sobering premise: shifting natural ecosystems, global travel patterns, and the potential misuse of AI are converging to create a more dangerous biosecurity landscape. Over the past 12 months, the teams have entered into more than 15 partnerships with government bodies, biosecurity organizations, and research groups as part of the effort.
“Breakthroughs like AlphaFold, IsoDDE, and AlphaGenome are radically shifting the balance. Instead of just reacting to natural outbreaks or safety risks, we can now design proactive defenses, accelerate therapeutics, and safeguard the global health ecosystem,” the post states.
The program is organized around three pillars: Prevent, Detect, and Respond.
Prevent focuses on ensuring AI models themselves are not misused. DeepMind has established a four-step safety process, threat modeling, evaluations, mitigations, and monitoring, applied to models like Gemini. In-house biologists and security experts test models against potential biological threats. The company is also adapting its SynthID watermarking technology to biology, which could help DNA synthesis providers screen for AI-generated biological sequences.
Detect draws on several AI systems. AlphaEvolve, a Gemini-powered coding agent, is being used to optimize algorithms for metagenomic sequencing data analysis, making it faster and cheaper to track diseases at scale. AlphaGenome, the genome-function model, is being explored for detecting and characterizing pathogens from sequence data, with the potential to identify novel patterns and emerging threats faster than conventional methods.
Respond involves providing trusted researchers with access to DeepMind’s latest AI systems to accelerate vaccine design and other countermeasures. Isomorphic Labs has established a dedicated unit to deploy its Drug Design Engine for medical countermeasures during novel outbreaks, working directly with governments and global health authorities.
The program aligns with DeepMind’s broader Frontier Safety Framework, which manages chemical, biological, radiological, and nuclear risks. Notably, the approach is explicitly collaborative: the company is sharing the framework openly and inviting partnerships with biosecurity labs, governments, and the scientific community.
The announcement comes at a time when the biosecurity implications of advanced AI are under intense scrutiny. Powerful protein-folding models and drug-design engines, originally developed for therapeutic purposes, have also raised dual-use concerns. DeepMind’s answer is to embed biosecurity directly into its model-development lifecycle while simultaneously building the infrastructure to respond faster when threats emerge.
Sources: “Our approach to bioresilience” (Google DeepMind Blog, July 16, 2026)