Disabled citizens and the future of AI

1. At a Glance

• The issue: AI is moving into government services, recruitment and healthcare. Disabled Indians are being consulted late or not at all, so these systems risk carrying disability bias and new access barriers [9].
• The opportunity: AI tools such as conversational screen readers, adaptive dashboards and real-time captioning can turn access into agency, when they are designed with lived experience at the centre [8].
• Legal anchor: In the Rajive Raturi judgment (Nov 2024), the Supreme Court held that accessibility is part of the fundamental right to life and dignity (Art. 21). It found India's accessibility rules non-binding, in conflict with the RPwD Act, 2016 [1][2][9].
• UPSC relevance: The topic joins up vulnerable sections (GS-II), AI governance and the digital divide (GS-III) and ethics of algorithmic fairness (GS-IV).

2. Why in the News

• An op-ed in The Hindu (6 Oct 2026, p. 9, by Navin Aade of NClude) argues that any AI system the government deploys should be tested for disability bias. It says training datasets need genuine disability representation, gathered with consent [9].
• The trigger is the Centre's push for about $200 billion in AI investment over the next two years, with data centres and power lines being built [9].
• Nearly two years after the Nov 2024 Raturi verdict, the petitioners are back in the Supreme Court because little has changed on mandatory standards [9].

3. Background & Evolution

• UNCRPD: The UN General Assembly adopted the Convention on 13 Dec 2006 and India ratified it on 1 Oct 2007. The RPwD Act, 2016 was passed to give effect to it [3].
• 2005: Rajive Raturi, a visually challenged person working with a human-rights organisation, filed a writ petition in the Supreme Court. He sought safe and accessible roads, public transport and other public facilities [2].
• 2016: The Rights of Persons with Disabilities Act, 2016 (Act No. 49 of 2016) was enacted, replacing the PwD Act, 1995 [3].
• 8 Nov 2024: The Supreme Court decided Rajive Raturi v. Union of India (2024 INSC 858). It directed the Centre to frame mandatory accessibility standards within three months [1][2].
• 7 Mar 2025: The Supreme Court adjourned the follow-up hearing of the Raturi PIL to this date [2].
• 30 Apr 2025: The Supreme Court held that "digital access is a right" and directed changes to KYC norms for persons with disabilities [4].
• 30 May 2025: DEPwD and ALIMCO held the "AI for Empowering PwDs" national conference in Bengaluru [5][6].
• Oct 2025: The UN published its analysis "Can AI create a fairer future for persons with disabilities?" [8].
• Oct 2026: The petitioners returned to the Supreme Court over non-compliance [9].

4. Core Static Facts

5. Multi-Dimensional Analysis

Legal / Constitutional

• Accessibility is now treated as an Art. 21 right to life and dignity. Rules that are merely advisory cannot meet the binding intent of the RPwD Act [1][9].
• The Apr 2025 ruling extends this to the digital sphere (e-KYC). By logical extension, AI-driven public interfaces are covered too [4].
• Gap: The Centre has not complied with the 3-month deadline, which is why the petitioners are back in court [9].

Social

• AI can give blind and low-vision users independence in everyday tasks such as reading documents, describing photographs and filling government forms that screen readers cannot open [9].
• Risk: If disabled people are excluded at the design stage, they are forced to adapt to systems built without them. The article calls AI "the newest such system" [9].

Scientific / Technological

• The relevant technologies include deep learning, robotics and computer vision applied to assistive tools [5].
• ALIMCO is adding AI components to next-generation prosthetics [7].
• Dataset bias: Training data under-represents disability. The fix proposed is consented, representative data [9].

Ethical / Governance

• The proposal is mandatory disability-bias audits before the state deploys any AI system [9].
• The UN stresses "nothing about us without us", meaning lived experience should be at the heart of design [8].

Economic

• The AI investment push of about $200 bn comes with physical infrastructure such as data centres and power lines. Inclusion has to be built in from the start, because retrofitting later costs more [9][2].
• Biased AI in recruitment can shut disabled people out of jobs [9].

Administrative

• Delivery is spread across DEPwD, ALIMCO and service-delivery ministries. The Unified Benefits Interface is meant to bring entitlements together in one place [5].

6. Recent Developments (last 12–18 months)

• 30 Apr 2025: The Supreme Court ruled that digital access is a right and ordered accessible KYC norms [4].
• May 2025: ALIMCO announced AI-integrated next-generation prosthetics [7].
• 30 May 2025: DEPwD and ALIMCO held the "AI for Empowering PwDs" conference in Bengaluru. DEPwD Secretary Rajesh Aggarwal said "a new era of inclusion has been ushered in by AI" [5][6].
• Oct 2025: UN News featured AI as a tool for a fairer future for persons with disabilities [8].
• Oct 2026: Raturi petitioners returned to the Supreme Court and the op-ed called for disability-bias testing of government AI [9].

7. Prelims Hooks

• Rajive Raturi v. Union of India is cited as 2024 INSC 858 and was decided on 8 Nov 2024 [1].
• The original Raturi writ petition was filed in 2005 [2].
• In the Raturi case the Supreme Court gave the Centre three months to frame mandatory accessibility standards [2].
• The Supreme Court held that accessibility rules based on guidelines and self-regulation do not meet the RPwD Act's mandate [2].
• The Supreme Court mandated a two-pronged approach: retrofit old infrastructure and make new infrastructure inclusive from the outset [2].
• The RPwD Act, 2016 is Act No. 49 of 2016 [3].
• The RPwD Act gives effect to the UNCRPD [3].
• "Digital access is a right" is the Supreme Court's Apr 2025 ruling on KYC for persons with disabilities [4].
• The "AI for Empowering PwDs" conference was held in Bengaluru on 30 May 2025 [5].
• That conference was organised by DEPwD with ALIMCO, not MeitY [5].
• Mission AI Accessibility and the National Disability Support AI Chat Bot were focus areas of the DEPwD conference [6].
• ALIMCO is the PSU making AI-integrated prosthetics [7].

8. Why a Court Order Has Still Not Become a Binding Rule

• The government's answer to accessibility has been to point to guidelines
• Rule 15 of the RPwD Rules, 2017 was amended in 2023 [10].
• The amended rule relied on documents such as the Harmonised Guidelines and Space Standards for Universal Accessibility in India, 2021, which were prepared by the Ministry of Housing and Urban Affairs [11].

• In Raturi, the Supreme Court said this approach was still self-regulation (each body checking itself, with no penalty). A guideline tells a builder what is good practice. A standard is a rule they must follow, and they can be held to account if they break it [2].

• The deadline was short, and it has long passed

• The court gave the Centre three months from 8 Nov 2024. That means early 2025 [1][2].
• The Centre was to draft the new rules with help from the Centre for Disability Studies at NALSAR University of Law, Hyderabad [2].

• Nearly two years later the petitioners are back in court because binding standards still do not exist [9].

• This delay matters for AI as well as for buildings

• The Apr 2025 KYC ruling made digital access a right [4]. Even so, there is no binding technical standard for testing a government app or AI system for accessibility.
• So an AI tool can be bought and launched with no legal test that it works for a blind or deaf citizen. The only option left is to go to court after the harm is done.

9. How an AI System Learns to Exclude Disabled People

• It learns from a past that already excluded them
• About 7 in 10 persons with disabilities around the world are outside the labour force. For people without disabilities the figure is about 4 in 10 [12].

• A hiring AI is trained on data about who was hired before. So it can learn that 'people like this are not hired' and repeat that pattern [12].

• It reads disability as a weakness, even when nobody tells it to

• AI résumé screeners ranked CVs that mentioned disability-related awards lower than CVs that were otherwise exactly the same [12].
• AI video-interview tools score eye contact and voice rhythm. This puts candidates with facial paralysis, hearing loss or neurodivergence (brains that work differently, such as autism or ADHD) at a disadvantage [12].

• None of these tools asks 'are you disabled?'. The bias comes in through features that look neutral.

• It keeps penalising people after they are hired

• AI systems that monitor work performance and enforce rigid schedules disadvantage employees who need reasonable accommodation (adjustments such as flexible hours or extra breaks) [12].

• Why this is a bigger risk when the state uses AI

• A biased private employer harms the people who apply to that one firm.
• A biased government system for recruitment, health triage or benefits applies the same error to every citizen who must use it, and those citizens have no other provider to turn to [9].

10. Why Disability-Specific AI Projects Do Not Solve the Problem

• India's AI effort for disabled people is mostly about assistive tools
• DEPwD and ALIMCO are working on AI prosthetics, an AI chatbot for disability support and Mission AI Accessibility [5][6][7].

• These projects build AI for disabled people. They are useful.

• But most of the risk comes from mainstream AI

• The op-ed's warning is about AI in general government services, recruitment and healthcare [9]. Those systems are run by other ministries, not by DEPwD.

• The 'AI for Empowering PwDs' conference was a DEPwD–ALIMCO event, not a MeitY one [5]. This shows that disability is still treated as one department's subject. It is not yet a test that every AI project must pass.

• What this means

• A good disability chatbot cannot fix a biased hiring algorithm in another department.
• Unless disability-bias checks are built into the main AI programme and into government buying rules, inclusion stays in a separate silo while the mainstream systems grow.

11. The Case Against Mandatory Bias Audits, and Why It Falls Short

• The strongest objection: audits will slow down India's AI push
• India is seeking about $200 bn of AI investment in two years [9]. A compulsory audit before every deployment adds cost and delay. Small Indian start-ups would feel this most.
• Disability is not one category. Blindness, deafness, autism and mobility loss each need different tests. So it is fair to ask what exactly an audit would check.

• Collecting disability data to train AI means collecting sensitive personal information. This raises real privacy risks.

• What the objection gets right

• A single, one-size audit would be shallow and could become a box-ticking exercise.

• Data collection must be consent-based, which is exactly what the op-ed itself demands [9].

• Why it still does not hold

• The Supreme Court has already decided that fixing access later costs more than building it in from the start. That is why it ordered new infrastructure to be inclusive by design [2]. The same logic applies to AI code just as it does to concrete.
• The EU, which is also trying to attract AI investment, still flags recruitment and performance monitoring as risk areas for disability discrimination in its AI Act [12]. Regulation and investment are not either-or.
• The answer is to match the test to the risk. Use strict audits for high-stakes uses such as hiring, health and welfare eligibility. Use lighter checks for low-risk tools. Do not drop the audit altogether.

12. Who Should Do What Next

• DEPwD: notify binding standards, including digital and AI
• Turn the Raturi order into enforceable rules under the RPwD Act. Do not refer back to guidelines again [2].

• Cover digital interfaces and AI systems in these rules, building on the Apr 2025 'digital access is a right' ruling [4].

• MeitY / IndiaAI Mission: make disability-bias testing a condition of government AI purchases

• Before a ministry deploys an AI system in recruitment, health or welfare, the system should be tested on disabled users. This is the op-ed's core demand [9].

• The EU AI Act gives a working model: it treats AI used in hiring and worker monitoring as an area of discrimination risk [12].

• Recruiting bodies and departments: keep a human in the loop

• The ILO recommends meaningful human oversight of AI decisions about jobs, plus training for HR staff on using AI ethically [12].

• A disabled candidate rejected by a machine should have the right to a human review.

• All agencies: involve disabled people in design, not only in feedback

• The UN's principle of 'nothing about us without us' [8] and the ILO's advice to include persons with disabilities throughout AI design and testing [12] both point the same way: put disabled people's organisations (groups run by disabled people themselves) on the teams that design and test the system.

13. Anchors for Answers

• Data: About 1.3 billion persons with disabilities worldwide. 7 in 10 are outside the labour force, against 4 in 10 for people without disabilities [12]
• Data: India seeks about $200 bn in AI investment over two years. This is the window in which inclusion is cheapest to build in [9]
• Report/Committee: ILO, 'AI's double-edged sword' (Oct 2025): AI résumé screeners ranked CVs with disability awards lower [12]. MoHUA Harmonised Guidelines and Space Standards for Universal Accessibility, 2021 [11]
• Law/Case: Rajive Raturi v. Union of India (2024 INSC 858), which placed accessibility under Art. 21 [1]. Apr 2025 SC ruling that digital access is a right [4]. RPwD Act, 2016 [3]. Rule 15 of the RPwD Rules, amended in 2023 [10]
• Comparison: EU AI Act: names recruitment and performance monitoring as areas where AI can discriminate against disabled people [12]
• Scheme: Accessible India Campaign (Sugamya Bharat Abhiyan), whose built and digital accessibility goals need binding standards behind them. DEPwD's Mission AI Accessibility [6]

14. Mains Relevance

• GS-II: Welfare schemes for vulnerable sections; mechanisms, laws and institutions for their protection; the Supreme Court's role.
• GS-III: Awareness in IT, AI and computers; inclusive growth.
• GS-IV: Ethics in public administration; fairness and accountability in algorithmic governance.
• Possible question stems: 1. "Accessibility is not a matter of charity but a facet of Article 21." Discuss in light of the Rajive Raturi judgment (2024). (15 marks) 2. As AI enters public service delivery, examine the risks of disability bias and suggest a governance framework for inclusive AI in India. (15 marks) 3. "Designing for disability at the outset is cheaper and fairer than retrofitting." Evaluate this with reference to India's digital and AI infrastructure push. (10 marks)

15. Related Topics to Study Next

• RPwD Act, 2016 and RPwD Rules, 2017: These are the statutory base and contain the rule struck down as non-mandatory.
• UNCRPD: This is the international treaty that the RPwD Act implements.
• Sugamya Bharat Abhiyan (Accessible India Campaign): This is the flagship scheme for built and digital accessibility.
• IndiaAI Mission and AI governance guidelines: This is the policy framework through which AI deployment happens.
• Digital Personal Data Protection Act, 2023: Consent-based data collection is central to building representative datasets.
• Algorithmic bias and Responsible AI: These ethics issues extend to caste, gender and language.
• Digital divide and e-governance (UMANG, DigiLocker): These are service channels that must stay accessible.
• Unique Disability ID (UDID): This is the identity backbone for disability entitlements and the UBI.

16. NCERT Concepts to Revise

• Rajive Raturi case (Polity): The central judgment: it made accessibility standards mandatory and is the legal anchor for the whole article.
• Rights of Persons with Disabilities Act 2016 (Polity): The statute the court said the guidelines contradict; it is the legal base for disability rights.
• Right to life and personal liberty (Polity): Article 21 is where the Supreme Court placed accessibility, reading dignity into the right to life.
• Accessible India Campaign (Polity): The flagship scheme for making buildings, transport and digital systems accessible, closely tied to the accessibility standards issue.
• Digital Personal Data Protection Act 2023 (Polity): Consent-based data collection is the proposed way to build representative disability datasets for AI.
• Digital divide (Geography): Shows how AI and digital services can leave disabled citizens behind without accessible design.

17. Common Errors / Trap Areas

• Petitioner and year: The petition was filed by Rajive Raturi in 2005, but the landmark judgment came in Nov 2024. Don't confuse the two dates [1][2].
• Wrong ministry: The AI-for-PwDs conference was run by DEPwD (Ministry of Social Justice & Empowerment) with ALIMCO, not MeitY [5].
• Rights basis: The Supreme Court located accessibility in Art. 21 (life and dignity), not only in Art. 14 or the Directive Principles [9].
• Act confusion: The current law is the RPwD Act, 2016. It replaced the PwD Act, 1995 [3].
• Two separate rulings: The Raturi (2024) case is about physical accessibility standards, while the Apr 2025 KYC ruling is about digital access. Don't merge them [2][4].