What can the Evo 2 genome AI do?

An AI foundation model trained on DNA is changing genome work

Evo 2 is a large genomics foundation model trained on an unprecedented corpus of DNA — reported at trillions of bases — and it can perform tasks that previously required specialized tools or expert curation. The system can identify genes, regulatory elements and splice sites, and it can generate short synthetic DNA sequences. Those capabilities position Evo 2 as both an analysis engine and an assistive design tool for synthetic biology and genomics research.

Why this is significant

• Scale: training on massive genomic datasets lets the model learn statistical patterns across many species and sequence contexts, improving generalization.
• Versatility: the model supports annotation (finding functional sites in genomes), prediction (inferring likely effects of variants), and generation (producing novel sequences with desired features).
• Accessibility: as an open-source resource, it can accelerate research across labs and sectors, from basic biology to biotechnology applications.

Opportunities and risks

• Opportunities:
• Faster gene and regulatory discovery across diverse organisms.
• Prototype design of short genetic parts for research and therapeutics.

• Tools for comparative genomics and evolutionary studies.

• Risks and safeguards:

• Potential to accelerate creation of harmful biological sequences if misused.
• Need for governance, access controls and biosafety review when models generate functional DNA.
• Ongoing validation is required: computational outputs must be tested experimentally before clinical or environmental use.

Evo 2 expands what AI can do in the life sciences, offering powerful new tools while underscoring the need for responsible deployment, experimental validation and ethical oversight.