Google DeepMind and Isomorphic Labs unveiled joint AI-based biosecurity program

Google DeepMind and Isomorphic Labs presented a joint approach to bioresilience on July 16, 2026 — a program designed to both prevent the misuse of their AI models for creating biological threats and provide scientists, authorities, and biosafety specialists with tools to protect against future disease outbreaks.

Why companies engaged with biosecurity

Google DeepMind explains: the biosecurity landscape is changing rapidly due to ecosystem shifts, increased global travel, and the risk of AI misuse. At the same time, artificial intelligence itself is called a critical tool for responding to these challenges — advanced models and scientific breakthroughs they enable should help society become more resilient to future outbreaks.

Work proceeds in two directions simultaneously: preventing models from being misused by bad actors while ensuring access to technologies for governments, scientists, biosecurity experts, and the companies' own teams.

What has been accomplished in a year

• Over the past 12 months, the companies have developed over 15 partnerships with government agencies, biosecurity organizations, and research groups
• Partnerships are aimed at preventing model misuse, detecting new outbreaks quickly, and responding to them promptly
• For models like Gemini, a four-step security process is applied: threat modeling, assessments, mitigation measures, and monitoring
• SynthID watermarking technology is being adapted for biology — it can help DNA synthesis providers filter out potentially dangerous AI-generated biological sequences

What AI developments underpin the program

At the core are scientific results the companies call breakthrough achievements: AlphaFold from Google DeepMind, which mapped the three-dimensional structures of nearly all known proteins; Isomorphic Labs' IsoDDE drug design engine (AI-powered Drug Design Engine), providing the precision needed for work with new biological systems; and AlphaGenome, shedding light on genome functions.

According to the companies, these systems enable a shift from simply responding to natural outbreaks and security risks toward proactive defense — accelerated search for therapeutic treatments and faster, more accurate protection of the global healthcare system. Google DeepMind and Isomorphic Labs provide models and agents based on them to trusted partners along three lines: prevention, detection, and threat response.

To work on risk prevention, the companies engage staff biologists and security specialists, as well as external partners — to understand potential threats, test models for resilience against them, and embed protective mechanisms.

What this means

Google DeepMind and Isomorphic Labs are formalizing what the AI industry has long warned about: the same models that accelerate drug development and genome understanding could theoretically be used for harm. A public bioresilience program is a major player's attempt to show that risk is manageable through partnerships with government and technical means like SynthID watermarks, not just declarations.

Frequently asked questions

What is

Google DeepMind and Isomorphic Labs' bioresilience program?

It is a joint initiative presented on July 16, 2026, combining measures to prevent AI models from being misused in biology and tools to help scientists and biosafety experts detect and respond to disease outbreaks.

How is SynthID connected to biosecurity?

SynthID watermarking technology, originally created to mark AI-generated content, is being adapted for biology: it can help DNA synthesis providers verify that ordered sequences were not generated by artificial intelligence and do not carry risk.