Announcing our partnership with the Republic of Korea

Source: DeepMind Date: 2026-04-27 URL: https://deepmind.google/blog/announcing-our-partnership-with-the-republic-of-korea/

Summary

Google DeepMind and South Korea’s Ministry of Science and ICT (MSIT) announced a partnership covering life sciences (AlphaFold, AlphaGenome, AlphaMissense), climate and energy (WeatherNext), and algorithm design (AlphaEvolve). An AI Campus will open in Seoul within DeepMind’s offices; a National AI for Science Center opens May 2026. Collaborating institutions include Seoul National University, KAIST, three MSIT AI Bio Innovation Hubs, and the Korean AI Safety Institute. The partnership includes internship pipelines for local AI talent development.

Implications

The science model portfolio — AlphaFold, AlphaGenome, AlphaEvolve, WeatherNext — is the substance, not the MOU. This isn’t a general AI partnership; it’s deployment of Google’s specific science tooling into Korean research infrastructure. SNU and KAIST getting hands-on access to AlphaGenome and AlphaMissense for Korean-population genomics datasets is a concrete research capability transfer, not a symbolic signing ceremony.

Korea’s AI Safety Institute as a named partner is the governance signal. Including the Korean AISI alongside MSIT means Google is building government safety research relationships in Korea on the same pattern as its UK AISI MOU. Two government AI safety institutes with formal DeepMind partnerships creates a template that other countries’ AI safety bodies will note — and potentially pressure their own governments to replicate.

National AI for Science Center opening May 2026 is the commitment with a hard deadline. Unlike “we will establish X” language, a named facility opening in five weeks is a concrete deliverable. Its existence and operational state will be observable — making this one of the more verifiable commitments in a partnership announcement.

Watch:

• Which Korean research outputs emerge from the AI Bio Innovation Hubs using AlphaFold/AlphaGenome — Korean genomic datasets have distinct population characteristics that could yield novel variant pathogenicity predictions via AlphaMissense
• Whether the Korean AISI collaboration produces joint safety publications, or remains a relationship without research output
• Whether this MOU structure (science tools + campus + safety institute + national center) becomes Google’s standard country-level AI partnership template following the India ANRF and UK government deals