National Strategy for Artificial Intelligence: #AIforAll (NITI Aayog discussion paper, June 2018)

The strategy has its origin in the Union Budget speech for 2018-19, which mandated NITI Aayog to establish a National Program on AI to guide research and development in new and emerging technologies. The paper describes a three-pronged approach: proof-of-concept AI projects with technology partners, a national strategy, and collaboration with experts and stakeholders.¹ It is a discussion paper, not a policy approved by the Cabinet or Parliament, and its foreword by NITI Aayog CEO Amitabh Kant says it is meant to be an "essential pre-read" for a later strategy. It states that execution and financial implications were "consciously avoided" because wider consultation was needed.¹

The paper frames India's opportunity around its scale and its development priorities. Its '#AIforAll' formula means technology leadership aimed at "the greater good": enhancing human capabilities, addressing shortages of skilled expertise, and building scalable solutions for emerging economies. Its technology stance is to "maximise the late-movers' advantage", adapting global advances to Indian needs rather than racing to pioneer them, while building research capacity for the long run.¹

Many of the later pieces of India's AI policy trace back to this document: a national compute facility, a public dataset platform, centres of excellence, an AI skilling push and a responsible-AI framework. NITI Aayog's own follow-up papers on responsible AI treat it as the "seminal document" of Indian AI policy, and a 2022 NITI Aayog paper credits it with coining the 'AI for All' benchmark for AI design and deployment.²³

NITI Aayog selected five sectors that it expected to benefit most from AI in meeting social needs, and where it thought government had to lead because private-sector effort "may neither be financially optimal nor efficient" alone:¹

Across these sectors the paper identified common barriers to adoption: a lack of broad-based expertise in AI research and application; no enabling data ecosystem and little access to "intelligent data"; high resource costs and low awareness among businesses; unclear privacy, security and ethics regulation, including the absence of formal anonymisation rules; and the absence of a collaborative approach. A sixth theme, an unattractive IP regime, appears in the later chapter on key challenges.¹

The paper's most distinctive proposal is a two-tier research system. It starts from an existing four-tier framework in the detailed project report of the Inter-Ministerial National Mission on Interdisciplinary Cyber Physical Systems (ICON, CROSS, CASTLE and CETIT) and argues for a "far more simplified and agile approach".¹

Centres of Research Excellence (CORE) would focus on fundamental research and next-generation AI architectures and platforms. The paper suggests starting at IISc, ISI and the top IITs and IIITs, with a hub-and-spoke relationship to other institutes and ties to AIIMS and TISS for multidisciplinary work. Possible research themes include sensory AI, physical AI, cognitive AI, general AI, learning from small datasets, new cryptographic and security algorithms, and explainable AI. It suggests funding of INR 50-100 crore per CORE and National AI Fellowships to retain Indian PhD students and attract the diaspora.¹

International Centres for Transformational AI (ICTAI) would be industry-led, application-focused institutions in the priority sectors. They would be public-private partnerships under a professionally managed society or Section 8 company called "ICTAI Inc.", with a majority of independent private-sector representation, which would select, fund and oversee them and could commission "moonshot" projects. The paper suggests seed funding of INR 200-500 crore per ICTAI for five years, shared between government and a corporate partner chosen by a "challenge method", and cites the French PRAIRIE Institute as a model.¹ An umbrella Centre for Studies on Technological Sustainability (CSTS), modelled on Singapore's CREATE or Innovate UK, would study pricing models, social penetration and international collaboration.¹

AIRAWAT (AI Research, Analytics and knoWledge Assimilation plaTform) would be a cloud platform with about 100 AI petaflops of capacity, similar to Japan's ABCI, connecting all COREs and ICTAIs through the National Knowledge Network. The paper also proposed a "CERN for AI", a supranational research effort, and "moonshot" projects such as advanced NLP for Indian languages.¹ NITI Aayog later published an AIRAWAT approach paper, and a Rs 7,500 crore Cabinet proposal for 20 AI adaptation centres, five CORE institutes and AIRAWAT was reported in 2019.⁴⁵

On adoption, the paper proposed that government act as a catalyst: opening social-sector datasets in machine-readable form, funding India-specific annotated datasets, using crowdsourcing and grand challenges for annotation, and setting up an AI database portal for collaborations. Its central market proposal is a National AI Marketplace (NAIM), with three modules: a data marketplace, a data-annotation marketplace and a deployable-model marketplace. The paper argues that informal data purchasing gives incumbents an oligopoly and prevents price discovery, and that a multi-stakeholder marketplace would reduce the time and cost of assembling data.¹ Other suggestions include an AI readiness index across states, a central-state shared fund for AI projects, incubation hubs and venture funds for AI startups, and government and PSUs acting as early adopters.¹

On skills, it cites a NASSCOM projection that 46% of the workforce would be in new or radically changed jobs by 2022 and a forecast gap of 200,000 data-analytics professionals by 2020. Its recommendations include tax holidays or CSR credit for new service jobs such as data annotation, AI and data-science training standards under the National Skills Qualification Framework, tax breaks for employer reskilling, credit-bearing MOOCs, bridge courses for non-computer-science postgraduates, AI modules in Atal Tinkering Labs, and a standing task force on AI-driven employment change.¹

The paper has a chapter on 'Responsible AI'. On fairness it notes that data can carry historical bias and suggests a use-case-based approach to identifying and reducing bias. On transparency it argues that opening the "black box" should aim at explainability rather than disclosure of source code, and it refers to DARPA's Explainable AI programme.¹

On privacy it welcomes the data protection work of the Srikrishna Committee, with its seven principles of informed consent, technology agnosticism, controller accountability, data minimisation, holistic application, deterrent penalties and structured enforcement. It adds sectoral regulation, benchmarking against the EU's GDPR, privacy-preserving research (differential privacy, privacy by design, multi-party computation), privacy impact assessments and public awareness campaigns.¹

For accountability it suggests shifting the debate from who is liable towards identifying which component failed, in the manner of aviation accident investigations. Its proposed framework has four elements: a negligence test rather than strict liability, safe harbours for developers who design, test and monitor properly, apportionment of damages in proportion to fault rather than joint and several liability, and an "actual harm" requirement so that suits cannot rest on speculative future damage. It proposes a consortium of Ethics Councils at every CORE and ICTAI. On intellectual property, it proposes a joint task force of the Ministry of Corporate Affairs and DIPP (now DPIIT) and IP facilitation centres, and notes that narrowly focused patent law may fit AI applications poorly.¹

The strategy was followed by a series of partial initiatives rather than a single national programme. On 7 September 2018 NITI Aayog, Intel and the Tata Institute of Fundamental Research announced a model ICTAI in Bengaluru, and a Maharashtra-linked centre focused on rural healthcare followed in November 2018. NITI Aayog also signed an AI partnership with Microsoft in 2018.³⁶ The National AI Portal (INDIAai) was launched on 30 May 2020.³ NITI Aayog's two-part approach document on responsible AI appeared in 2021 and a use-case paper on facial recognition in 2022, treating the 2018 strategy as the foundation.²

The Union Cabinet approved the IndiaAI Mission with an outlay of Rs 10,371.92 crore on 7 March 2024, which put into effect several of the 2018 themes: a public compute facility, a public datasets platform, India-specific foundation models and AI skilling.³⁸ In September 2025, when NITI Aayog published its 'AI for Viksit Bharat' roadmap, MediaNama noted that the 2018 strategy had seen "little follow-up" other than the IndiaAI Mission and that India still had no Cabinet-endorsed national AI policy.⁷

Criticism and limits. The paper called its own recommendations "descriptive rather than prescriptive by design" and avoided funding targets, which left the CORE and ICTAI structure without a budget line. The ICTAI idea was only partly realised through industry partnerships, and the paper preceded India's data protection law, which was enacted in 2023.¹³ In 2025 the Future of India Foundation's Ruchi Gupta contrasted the 2018 strategy's "public-interest framing" with the more industry-focused 2025 roadmap and called for a Cabinet-endorsed national AI policy.⁷