Inclusivity in AI foundational, not optional: UNESCO's Tim Curtis

Inclusivity in AI is not optional but foundational for a country as vast and diverse as India, where systems must be designed from the start to reflect the nation's immense social, economic, and linguistic diversity, UNESCO South Asia Regional Office Director Tim Curtis said on Tuesday.

Curtis, speaking at the 5th and final Stakeholder Consultation on the AI Readiness Assessment Methodology (RAM) in the national capital, warned that if AI tools are not intentionally built to serve the full spectrum of users, including marginalized communities, they risk leaving millions behind.

''In a country as vast and complex as India, (with) not only 1.4 billion people, but with thousands of languages and immense social, economic and regional diversity, inclusivity in AI is not optional, it is foundational,'' Curtis said.

He pointed out that AI systems trained primarily on English or urban data can struggle with local languages, dialects, and contexts, leading to the exclusion of vast populations from digital services.

These are not just technical bugs, he explained, but symptoms of deeper structural exclusions in data collection and prioritization.

He stressed that inclusivity in AI means making deliberate choices about who is represented in datasets, who participates in development, and whose needs drive deployment.

Curtis highlighted the importance of ''ethics by design,'' arguing that responsible AI must be built on values as much as on functionality.

''If we want AI systems to be inclusive, we must design them that way from the start, encompassing not just functionality, but essential values. And when we don't, we begin to see consequences, not as abstract risks, but as real-world limitations,'' Curtis said.

Risks such as privacy violations, disinformation, and bias can arise from algorithmic decisions made without transparency or oversight, he said.