"Designing for disability at the outset is cheaper and fairer than retrofitting." Evaluate this with reference to India's digital and AI infrastructure push.

The RPwD Act, 2016 already uses design-first logic. New buildings cannot get approval unless they follow accessibility rules (Sec. 44), and existing public buildings were given five years to retrofit (Sec. 45) [2]. India is now building digital public infrastructure and AI at scale, and on that front the claim largely holds, with some caveats.

Why it is cheaper

• Retrofitting has a poor track record: a NALSAR-CDS assessment ordered in Rajive Raturi found widespread accessibility gaps in buildings, transport and ICT, even after statutory deadlines [1].
• Fixing digital exclusion later means court-ordered redesign: in Pragya Prasun (2025), the SC had to order digital KYC changes because liveness checks shut out acid-attack survivors and blind users [3].
• AI multiplies early errors: models trained on past hiring data repeat past exclusion. Résumé screeners ranked CVs with disability-related awards lower than identical CVs [5]. Correcting a state system after deployment means retraining and re-auditing it at national scale.

Why it is fairer

• Rights, not charity: Raturi rejected accessibility rules that left compliance to self-regulation through guidelines and ordered mandatory standards within three months [1]. Pragya Prasun held that digital access is part of Art. 21 [3].
• No exit option: a biased government platform affects every citizen who has to use it.
• Co-design: the ILO recommends involving persons with disabilities in AI design and testing, with human oversight of decisions [5].

Limits of the claim

• Upfront cost and complexity: blindness, deafness and autism each need different tests. Blanket audits could turn into box-ticking and weigh heavily on start-ups.
• Privacy: representative datasets need sensitive data, so they must be collected with consent.
• Legacy stock still needs retrofitting: Rule 15 still relies on guidelines such as the 2021 Harmonised Guidelines [6].

Way forward

• Notify binding accessibility standards that cover digital interfaces and AI [1][3].
• Make disability-bias testing a condition of government AI procurement for high-stakes uses (hiring, health, welfare), with lighter checks for low-risk tools.
• Move beyond DEPwD's assistive initiatives such as Mission AI Accessibility and the AI chatbot [4] so that every ministry's AI is held to the same standard.

Overall, court records and labour-market evidence show that designing for disability from the start is cheaper and fairer. It is not free, though, and it cannot replace retrofitting older infrastructure. A risk-based, co-designed approach would protect Art. 21 dignity, meet India's UNCRPD commitments and serve SDG 10's promise to leave no one behind.