Gates Foundation Forms 60-Partner AI Coalition to Build Language Data Sets for 3 Billion People

Gates Foundation Unites 60 Organizations to Address Language Gaps in AI

The Gates Foundation announced a global AI coalition Monday, bringing together 60 partners including Anthropic, Google, OpenAI Foundation, Microsoft, and the World Bank Group to build more representative language data sets for AI1

1

5

5

. The coalition targets reaching more than 3 billion people over five years by making AI tools accessible in underrepresented languages currently shut out from the technology2

2

3

3

. Bill Gates told CBS News the foundation initially hoped to bring together 10 organizations but ultimately secured participation from 60, calling AI "almost an evolutionary event" that differs fundamentally from previous technologies5

5

.

Why Language Representation Matters for Global Inequality

The initiative follows the foundation's Goalkeepers report, where it committed $1 billion toward AI-focused efforts to improve health outcomes, upgrade educational tools, and inform small farmers' practices worldwide1

1

4

4

. Unrepresentative language data creates dangerous mistranslations that could impact lives. The report warned that AI models might mistranslate a pregnant Malawi woman saying her "water has broken" into the direct English that she'd "thrown away water," potentially causing medical emergencies2

2

3

3

. Gates Foundation CEO Mark Suzman emphasized this work must continue "full speed ahead" even as some AI companies urge the industry to slow down development of advanced models, arguing governments should regulate impacts on cybersecurity and children's development while extending AI's humanitarian applications1

1

2

2

.

Addressing the Original Sin of Internet-Trained AI Models

Mozilla Data Collective CEO E.M. Lewis-Jong called the "original sin" of AI the fact that many tools were trained on language data scraped from the internet without cultural context or explicit consent1

1

3

3

. "The internet is not a representative space," Lewis-Jong said. "Why would you think that you were going to get a culturally diverse and representative system out of something that was predominantly trained on Reddit?"2

2

4

4

. The Mozilla Foundation subsidiary now operates a data-sharing platform that helps communities upload cultural and linguistic data sets on their own terms, ensuring data sovereignty and ethical data collection practices1

1

2

2

.

Four Pillars of the Coalition's Approach

Participating organizations will pool resources and expertise across four critical areas: building shared and safe open-language data infrastructure, developing benchmarks to track progress, turning language data into AI models and applications accessible to developers, and ensuring responsible deployment while protecting privacy, consent, and data sovereignty5

5

. The coalition will jointly develop its governance framework, organizational structure, and key workstreams over the next year5

5

. A secretariat will track each signatory's commitments, and the Gates Foundation may nudge partners to fill larger gaps when necessary, according to Mark Suzman1

1

3

3

.

Google's Project Vaani Tackles Dialect Diversity in India

Google has been funding Project Vaani, an effort to collect more than 150,000 hours of audio across every district in India1

1

4

4

. Google senior vice president James Manyika explained the push underscores the need to gather speech data on dialects within languages, working with local partners to record speech in the field2

2

3

3

. This approach highlights how the coalition addresses not just language representation but the nuanced variations within languages that affect AI performance in real-world applications across health, education, and agriculture sectors.

Anthropic Acknowledges Gaps in African Language Support

Anthropic was already working with the Gates Foundation to speed up vaccine developments and bolster its chatbot's data set of local crops before joining the coalition1

1

4

4

. Elizabeth Kelly, the head of beneficial deployments at Anthropic, acknowledged the company's products "lag in many African languages in particular"2

2

3

3

. "We're acutely aware that we can't achieve any of the benefits we want to see in terms of improving patient outcomes or improving literacy and numeracy for students across the globe unless we actually get this language piece right," Kelly told the Associated Press1

1

4

4

. This acknowledgment from Anthropic, whose CEO recently published an essay calling for industrywide cooperation on decelerating advancements, demonstrates the tension between rapid AI development and ensuring inclusive access2

2

3

3

.

Balancing AI Development Speed with Humanitarian Applications

At the Goalkeepers event during UN week, Bill Gates told political commentator Van Jones that "increased generosity" and "the smart use of AI" together could accelerate the fight against inequality2

2

. Gates acknowledged concerns about AI risks, stating "We failed to build in the right human values and ways of keeping our moral code controlling what these AIs do," while emphasizing the need to minimize risks like bioterrorism and job loss while accelerating benefits in education and health5

5

. Mark Suzman reinforced this position, telling the Associated Press that even if AI development were frozen, building these language sets remains critical for making existing tools usable1

1

2

2

. The coalition's formation comes as 26 of the world's 34 wealthiest nations decreased development aid to poorer countries last year, making the coordination of AI resources for humanitarian purposes increasingly urgent2

2

.