AI accessibility: Use India’s AI Summit to define it clearly and set a mandate for the world to adopt

In India, the Rights of Persons with Disabilities (RPwD) Act of 2016 provides a robust legal framework for persons with disabilities (PwDs). But it defines accessibility primarily in negative terms, with a ‘barrier’ taken as its denial. To ensure that “AI for All," the central theme of this year’s summit, becomes more than a slogan, we must construct a three-tiered definition of AI accessibility rooted in legal and disability jurisprudence.

The definitional gap: Under the RPwD Act, a ‘barrier’ is defined broadly as any factor , be it “communicational, cultural, economic, environmental, institutional, political, social, attitudinal or structural," that hampers the full participation of PwDs. By extension, Indian law has treated accessibility as the mere absence of these obstacles.

But AI is not a static physical structure like a ramp. It is a recursive form of software that evolves. If the summit is to set a global benchmark, it must go beyond calls to remove barriers and mandate what AI needs to be from its very inception.

For AI, accessibility must be understood as a spectrum of time and utility, operating at three distinct levels.

First, instant stage accessibility: This is an entry threshold that covers basic requirements for participation. It means ensuring that foundational models and interfaces are compatible with assistive technologies from day one. For instance, in the context of the IndiaAI Mission’s focus on multilingualism, we should ensure that speech-to-text tools for regional languages account for the diverse vocal patterns of PwDs. This is the ‘right to entry’ aspect.

Second, medium-term accessibility: For this, AI must move from mere compatibility to ‘active facilitation’. Accessibility at this stage is defined by convenience. AI should deploy features that ease the access of PwDs to goods or services compared to traditional means. Whether it is an AI-driven vision assistant to navigate government portals or a predictive interface to reduce motor load, we must focus on enhancing AI-interaction quality.

Third, long-term accessibility: This is the ultimate goal. Accessibility must translate into systemic ‘inclusion’. This occurs when the recursive nature of AI is used to eliminate historical biases. At this level, AI does not just ‘assist’ PwDs; it restructures the environment to ensure their participation as equals in society and the workforce.

The Raturi framework: A pivotal moment in Indian accessibility jurisprudence was the Rajive Raturi vs Union of India judgement of 2024. The Supreme Court made a sharp distinction between accessibility and reasonable accommodation. It characterized accessibility as an ex-ante obligation: i.e. a proactive and universal mandate that must exist the moment an entity or service is created.

Hence, ex-ante accessibility must be baked into the very training sets and ‘model cards’ of India’s sovereign AI projects. We mandatorily need universal designs to ensure that AI tools are usable by as many people as possible right from their initial deployment. Corrective patches that can be applied later will not do, as this is a legal and constitutional must.

Conversely, reasonable accommodation is an ex-post duty: i.e. an individualized adjustment made after a specific barrier is identified for a particular person. This is where AI’s greatest strength, its recursivity, comes into play. Because AI can learn and adapt, it can fulfil customized duties that physical infrastructure cannot.

If a universal interface fails to provide a tailored experience to a specific user with a unique combination of impairments, AI can be prompted or fine-tuned at the edge to deliver a more effective solution. The recursive nature of AI lets us bridge the gap between universal standards and individual needs in real-time.

Hope at the forthcoming summit: Scheduled next month, the India AI Summit 2026 will present us with an opportunity to harmonize technology with the legal jurisprudence of disability on accessibility. We must move away from the negative definition of removing barriers and towards an affirmative national AI accessibility standard, one that we could nudge the rest of the world to adopt.

This standard must recognize that while developers have an ex-ante duty to build accessible models, the AI model or tool itself must be trained to recognize when it needs to provide ex-post reasonable accommodation. The recursive nature of these models enables a self-correcting accessibility loop that was impossible in the era of brick-and-mortar infrastructure.

Defining accessibility for AI is not merely a technical challenge. It is imperative from the perspective of fundamental rights. By adopting the Rajive Raturi framework, the India AI Summit could help ensure that the next generation of technology is inclusive by design (ex-ante) and adaptive by nature (ex-post).

If India is to lead the Global South in the AI revolution, we must prove that our own code respects the dignity of all citizens. Our goal should be to ensure that in the automated future, the ‘barrier-free’ environment promised by our legal framework is finally realized in our algorithms.

The author is a lawyer and the founder of Politics and Disability Forum.