Google DeepMind has launched a programme to turn AI on biological threats. The London lab and its sister company Isomorphic Labs call it bioresilience, and set it out in a joint blog post.
The aim is twofold. DeepMind wants to stop people misusing its models, and to help governments, scientists and biosecurity groups use AI to prevent, detect and respond to outbreaks. It says it has built more than 15 such partnerships over the past year.
What the programme covers
On prevention, DeepMind runs a four-step safety process: threat modelling, evaluations, mitigations and monitoring. It is also adapting its SynthID watermarking to biology, so DNA-synthesis providers could screen for risky AI-generated sequences.
For detection, the company is using its AlphaEvolve agent to speed up the algorithms behind pathogen sequencing. That, it says, makes tracking new outbreaks faster and cheaper.
