DeepMind and Isomorphic Labs turn frontier biology models toward biosecurity defense

Google DeepMind and Isomorphic Labs said on July 16 they have formed more than 15 partnerships with government bodies, biosecurity organizations and research groups to redirect their frontier biology models toward biological-threat defense. The two labs call the approach “bioresilience” and say the partnerships came together over the past year.

In a joint post, the two labs organized the work around three areas: preventing, detecting and responding to biological threats. The argument is that the same AI systems built to advance biology can also guard against its misuse.

The post names specific tools repurposed for the effort. They include AlphaFold, the protein-structure prediction model; Isomorphic Labs’ Drug Design Engine; AlphaGenome, which analyzes how genomes function; a biology-adapted version of the SynthID watermarking system; and AlphaEvolve applied to metagenomic sequencing, a method for screening genetic material in environmental samples.

DeepMind framed the effort as a proactive-defense complement to its existing frontier-safety work, which aims to stop AI models from being used to help design biological weapons. Isomorphic Labs is the drug-discovery company spun out of DeepMind.

The post is a first-party account. No independent party has verified the tally of partnerships or specified what each one produces, and DeepMind did not name the government bodies involved or say how much of the work has moved past planning.

For a field where one model can speed a cure or a threat, the test is whether these partnerships ship defensive tools fast enough to matter.