Parliament Question: Strategy for Artificial Intelligence

GOVERNMENT OF INDIA MINISTRY OF ELECTRONICS AND INFORMATION TECHNOLOGY LOK SABHA STARRED QUESTION NO. *158 TO BE ANSWERED ON: 30.07.2025 STRATEGY FOR ARTIFICIAL INTELLIGENCE *158. DR. D RAVI KUMAR: Will the Minister of ELECTRONICS AND INFORMATION TECHNOLOGY be pleased to state: (a) the details of steps taken by the Government to formulate a clear national strategy for Artificial Intelligence (AI) development that ensures accountability, safety and fairness while prioritising human rights and privacy; (b) the manner in which the Government is planning to ensure that the proposed AI Safety Institute adopts a human rights-based approach with a focus on openness, innovation and inclusivity while safeguarding people’s data and privacy; (c) the details of measures being considered by the Government to strengthen data protection and cybersecurity regulations to prevent misuse including providing citizens with the right to opt out of their data being used for training AI models given the reliance of AI on personal data; and (d) the manner in which the Government is likely to address concerns about the use of AI in automated decision-making, especially in sensitive areas like policing, access to essential services and elections while ensuring that AI regulations do not suppress free speech or innovation? ANSWER MINISTER FOR ELECTRONICS AND INFORMATION TECHNOLOGY (SHRI ASHWINI VAISHNAW) (a) to (d): A Statement is laid on the Table of the House.STATEMENT REFERRED TO IN THE REPLY TO LOK SABHA STARRED QUESTION NO. *158 FOR 30.07.2025 REGARDING “STRATEGY FOR ARTIFICIAL INTELLIGENCE” …….. In line with Prime Minister Shri Narendra Modi's vision, the Government is making technology accessible to all. The focus is on democratizing the development and use of Artificial Intelligence (AI) for real-world problems, ultimately improving lives across various sectors. IndiaAI mission includes targeted interventions that, inter alia, ensure accountability, safety, fairness, and the protection of human rights and privacy. Development of India’s Foundational Models ● Developing indigenous Large Language Models/Small language Models on Indian data sets ● Models so trained will ensure a better understanding of local contexts, dialects, and cultural nuances ● Target areas: Healthcare, education, agriculture, climate, and governance ● Four (4) startups - Sarvam AI, Soket AI, Gnani AI, and Gan AI—have been selected (so far) to build India’s foundation models (Details in Annexure -1) The models developed under IndiaAI mission will be made open source to help other startups build applications, especially India specific ones, on top of it. IndiaAI Compute Capacity: Establishing high-end AI infrastructure access of GPUs to support safe and scalable AI research and innovation Aspect Details Total GPUs Provisioned 34,381 GPUs through the IndiaAI Compute Portal Academia, MSMEs, startups, research community, Beneficiaries government bodies, etc GPUs provided at discounted rates with 40% cost Pricing Support supported by the Government Around ₹67 per GPU-hour (less than 1 USD), which is Average Price on Portal one-third of the global average Nvidia H100, H200, B200; Intel Gaudi 2 & 3; AMD Types of GPUs Available MI300X, etc IndiaAI Datasets Platform (AIKosh) Ensuring inclusive AI development by offering curated datasets across sectors such as health, agriculture, and education, with safeguards for data privacy ● India-specific 1000+ datasets and 208 AI models are available on the platform● Examples of datasets - Farmer query data from Kisan Call Centres, geological data from states, clinical, imaging, & pathology data to support AI-based diagnosis of brain lesions ● Small AI models are also available on platform; For eg. Text-to-Speech (TTS) models in Indian languages like Bengali, Gujarati, Kannada, Malayalam Development of AI based applications ● Supporting AI startups solving public interest problems in governance, health, climate, etc., while following ethical design frameworks. ● Currently 30 applications currently being funded under the Mission (details in Annexure- 2) Support to AI-based startups in India IndiaAI Startups Global program is launched with Station F & HEC Paris, to help 10 Indian AI startups expand into Europe; Currently they are undergoing training at Station F. These include startups working in areas like responsible AI, and AI in cybersecurity, among others (Details in Annexure -3) Among these are startups like PrivaSapien Technologies, working on privacy-enhancing AI, and Secure Blink focused on AI-driven cybersecurity. Safe and trusted AI ● IndiaAI Safety Institute for AI safety, security, and developing trust with diverse stakeholders ● Diverse stakeholders including academia, startups, industry & government bodies engage through a hub-and-spoke model ● Tools and frameworks are being built around Responsible AI, with a focus on 10 themes ● 8 Responsible AI Projects underway covering Machine Unlearning, AI Bias Mitigation, Privacy-Enhancing Tools, AI Governance Testing, etc. ● More projects on key themes like watermarking, ethical AI, risk assessment, stress testing, and deepfake detection tools are also being evaluated India is also hosting the Al Impact Summit in February 2026, building on its role as co chair of the Al Action Summit and continuing its leadership in shaping global AI discussions. Legal Framework for Mitigating AI-Related Risks Government is cognizant of the risks posed by the usage of AI in automated decision-making, such as Hallucination, bias, misinformation, and deepfakes. The legal framework to ensure safety in this regard is as follows: Information Technology Act, 2000: ● Sections 66C (Punishment for identity theft) deals with misinformation, deepfakes, cheating by personation or identity theft.● Section 66D of the IT Act criminalizes the use of computer resources for cheating by personation. ● Section 66E prescribes the punishment for capturing and publishing or transmitting the image of a private area of any person without his or her consent. ● Section 67A and 67B make publishing or transmitting obscene material for instance, which could be generated by using deepfake technology a punishable offence. Legal provisions under Bharatiya Nyay Sanhita, 2023: ● Section 111 of the BNS punishes the commission of any continuing unlawful activity including economic offence, cyber-crimes, by any person or a group of persons, either as a member of an organised crime syndicate or on behalf of such syndicate. ● Several other sections under the BNS also deal with cyber-crimes like cheating or cheating by personation such as sections 318 (Cheating), 319 (cheating by personation), 353 (public mischief), 356 (defamation). Digital Personal Data Protection Act, 2023 ● It casts obligations on Data Fiduciaries to safeguard digital personal data, holding them accountable, while also ensuring the rights and duties of Data Principals. ● It empowers citizens with control over their data, ensuring consent-based usage and accountability of data fiduciaries. Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021 (“IT Rules, 2021”) ● The IT Rules cast specific legal obligations on intermediaries, including social media intermediaries and platforms, to ensure their accountability towards safe and trusted internet. ● This includes expeditious action needed towards the removal of the prohibited misinformation, patently false information and deepfakes. ● In case of failure of the intermediaries to observe the legal obligations, they are liable for consequential action or prosecution as provided under the extant laws. ● Under the IT Rules 2021, there is a provisions grievance redressal mechanism by the intermediaries which inter alia provides 24 hours timelines for any grievances relating to morphed or artificially generated images affecting the victim. ● If not satisfied with the grievance redressal, aggrieved persons can approach the Grievance Appellate Committee. ● The Ministry of Home Affairs has launched a dedicated portal to report cybercrimes [cybercrime.gov.in] and has also started a toll-free number 1930. Techno-legal approach to regulate AI A key pillar of India’s AI strategy is its balanced and pragmatic techno-legal approach to regulation: ● India is combining legal safeguards with technological solutions.● The Government is funding R&D projects at premier institutions like IITs to build AI tools for deepfake detection, privacy enhancement, and cybersecurity. ● This reflects India’s belief that effective AI governance must go beyond just legislation and be backed by innovative technical interventions. ● This distinctive model ensures that innovation is not hampered ******* Annexure 1 Details of projects approved for development of India’s Foundation Models: a. Sarvam AI has been selected to build India's Sovereign LLM Ecosystem, developing an open source 120 billion parameter AI model to enhance governance and public service access through use cases like "2047: Citizen Connect" and "AI4Pragati". b. Soket AI is working to develop India's first open-source 120 billion parameter foundation model optimized for the country's linguistic diversity, targeting sectors such as defense, healthcare, and education. c. Gnani AI is working to build a 14 billion parameter Voice AI foundation model delivering multilingual, real-time speech processing with advanced reasoning capabilities. d. Gan AI is working to create a 70 billion parameter multilingual foundation model targeting "Superhuman TTS (text-to-speech)" capabilities to surpass current global leaders.Annexure 2 Details of theme wise projects selected under IndiaAI Application Development Initiative: • Climate Change and disaster management: 4 applications have moved to the next stage under this theme with 1 idea and 3 prototypes. Indicative Title of the Project and Description Team Solution problem Name stage statement Early Warning Heatwave Resilience: Integrating AI-based ARTPARK Idea Systems for Advanced Forecasting for Extreme Heat Events: Adverse and Integrate AI-based climate forecasting with Sudden health data to develop a targeted risk Climate classification system for heat-health impacts Change Multi-Hazard Multi-Hazard Susceptibility Mapping: hazard Jarbits Pvt Prototype Susceptibility management system that integrates four modules: Ltd Mapping Hazard Prone Area Database, Susceptibility Mapping, Threat Assessment & Mitigation Planning, and Monitoring & Assessment. Solutions that DeepFlood: Rapid flood inundation mapping Indian Prototype correspond to using Vision Transformers and Satellite Data: Institute of both indicative advanced flood inundation mapping tool using Technology problem SAR data and deep learning models for real-time, Delhi statements automated flood detection Others: Moskeet : Climate impact on mosquito-borne Moskeet Prototype Climate and diseases: AI-powered platform integrates real- Outbreak of time climate data with mosquito tracking to Diseases predict and prevent disease outbreaks. • Healthcare: 8 applications have moved to the next stage under this theme with 4 prototype and 4 existing solutions. Indicative Title of the Project and Description Team Name Solution problem Stage statement Early disease End to End AI Cloud Platform for Radiology Endimension Existing detection using Diagnosis: AI cloud platform for radiology Technology Solution diagnosis, integrating innovative technologiesIndicative Title of the Project and Description Team Name Solution problem Stage statement AI-enhanced such as CV, GenAI, NLP, Dicom, mobile, and Private X-rays cloud computing Limited NIDAAN (iNtegrated lung health screening & Qure.ai Existing tuberculosIs Detection through Ai At National Technologies Solution scale): qXR, an advanced AI tool for interpreting Private chest X-rays (CXR), detects & localizes 30+ Limited findings. AI/ML enabled MafPro device platform for cancer SIAMAF Prototype staging, localization, and margins. MafPro Healthcare handheld detector provides a radiation-free, non- invasive, safe and cost-effective solution that can reliably detect and adequately evaluate metastases in lymph nodes using AI/ML based algorithms Strengthening Impactful AI solution, preventing preventable Artelus Existing ophthalmology blindness, for socio-economic transformation: Solution outcomes DRISTi AI is a CE-marked, device-agnostic using AI for retinal screening solution detecting 16 DR-related detection and abnormalities and 20 vision-threatening management conditions. of conditions VoxelBox: SaaS platform that integrates BrainSightAI Prototype advanced computational neuroscience, AI, and 3D visualization to provide clinicians with deeper insights into brain structure and function. Development of Ocellux: An AI-Based Solution Anself Prototype for Enhanced Early Diabetic Eye Screening in Dynamics INDIA: smartphone-based retinal imaging device Pvt. Ltd. designed for capturing high-resolution images of the retina and anterior segment Others: AI Powered Wearable Technology for Detection Team Existing Patients & Diagnosis of Musculoskeletal Joint Health Pain: Pheezee Solution MonitoringIndicative Title of the Project and Description Team Name Solution problem Stage statement non-invasive device that enables physiotherapists to assess joint health with precision. Others: Revolutionizing healthcare using doctor-led AI: Jivi Health Prototype Enabling AI-powered personal doctor available 24x7 and Private Access to free of cost, designed to assist people when they Limited Healthcare are sick, monitor their health through wearables, and act as a health coach to prevent diseases. • Agriculture: 8 applications have moved to the next stage under this theme with 1 prototype and 7 existing solutions. Indicative Title of the Project and Description Team Name Solution problem stage statement Soil Testing Krishi Sah‘AI’yak - Farming Co-pilot: AI- Samagra Existing and Advisory powered conversational assistant providing Development Solution personalized agricultural advisory in multiple Associates Pvt. Indic languages. Ltd KissanCopilot Multilingual Multimodal Titodi Infotech Existing personalized AI Assistant for small holder Pvt Ltd Solution farmers: Powered by Dhenu's vertical LLM for agriculture, provides personalized agricultural advisory services to smallholder farmers. Rapid, chemical-free soil testing solution using Neoperk Existing NIR Spectroscopy & AI/ML models: End-to- Technologies Solution end soil testing approach leverages an offline Pvt. Ltd. app, AI-powered spectral device, and ML models to deliver timely, chemical-free soil analysis. Others: MRV: in-house developed MRV Varaha Climate Existing Carbon (Measurement, Reporting, and Verification) AG Private Solution Credits for technology designed to ensure the integrity and Limited Soil Health quality of generated carbon credits.Indicative Title of the Project and Description Team Name Solution problem stage statement Water and Kadalcompass - We Amplify the voice of water Innogle Existing Fish using Hydro-AI,Underwater com,IoUT with Technologies Solution Monitoring Sensor Network: 5G-enabled device that Pvt LTD transforms fishing practices using AI, IoUT, and advanced communication technologies AI-Driven Live Fish Logistics and Mortality SNRAS PVT Existing Reduction in Aquaculture: AI-driven solution LTD Solution optimizes aquaculture operations by monitoring water quality, predicting potential issues, and providing prescriptive recommendations to ensure fish health and reduce mortality. Quality of AI-Deep tech driven quality assessment of RootsGoods Existing produce maize: Traceability of the maize quality from (OPC) Pvt. Ltd. Solution farm to processing units provides unique and important value for the entire product chain in the market Q-scan: AI-powered IoT solution that captures Qzense Labs Prototype critical data throughout the fresh produce Pvt Ltd lifecycle, providing real-time analysis of food quality and actionable insights to sellers. • Governance: 4 applications have moved to the next stage under this theme with 1 prototype and 3 mature solutions. Indicative Title of the Project and Description Team Name Solution problem Stage statement Public service ConvoZen.AI: AI-powered conversational Nobroker Existing access platform that automates customer engagement Technologies Solution across channels like chat, voice, email, and social Solutions media, offering fast, personalized, and Pvt. Ltd multilingual support.Indicative Title of the Project and Description Team Name Solution problem Stage statement AI contact center: AI-powered technologies like Devnagri AI Existing machine translation, NLP, ASR and TTS, and Pvt. Ltd Solution multilingual voice recognition to enhance governance by improving communication and accessibility across India’s linguistic diversity. Grievance Adalat AI: AI solutions for Courts: legal tech Legsys.AI Prototype Redressal and solution leverages AI-driven transcription to Technology Legal Support automate court processes, reducing India's Foundation overwhelming judicial delays. Others: Gov.Civis.Vote: Leverages Civis' proven public Civic Existing Improving consultation framework and integrates advanced Innovation Solution Internal technologies to create a unified platform for Foundation Administration governments. • Assistive technologies for learning disabilities: 6 applications have moved to the next stage under this theme with 2 ideas, 2 prototype and 2 mature solutions. Indicative Title of the Project and Description Team Name Solution problem Stage statement Early Jiveesha: AI-powered platform for diagnosing Daira Edtech Idea Identification Specific Learning Disabilities using advanced Private for SLDs software and a plug-and-play device. Limited ScreenPlay - a digital game-based screening tool Kidaura Existing or autism and related disorders: AI-driven, game- Innovations Solution based platform designed for early identification Pvt Ltd and intervention of Autism Spectrum Disorder in children aged 3-6. Enhanced Readabled (Online Dyslexia Training): web-based Hermit Labs Existing Multimedia application designed to help children with Pvt. Ltd. Solution Accessibility dyslexia improve phonetic awareness through tools interactive exercises. Voice fusion AI: adaptive voice cloning platform PerpetualBlock Prototype designed to empower speech-impaired individuals TechnologiesIndicative Title of the Project and Description Team Name Solution problem Stage statement to express themselves through emotionally Private connected, lifelike voice outputs, Limited Solutions Adaptive Learning and Detection for SLDs: IITK- Prototype that address advanced AI techniques to detect Specific MadhavLab both Learning Disabilities (SLDs) such as dyslexia, indicative dysgraphia, and dyscalculia. problem Special Educator AI: AI-driven system designed to Viranc Idea statements address India’s shortage of special educators and Infotech support children with Specific Learning Disorders. Annexure 3 Startups supported under IndiaAI Startups Global program launched with Station F & HEC Paris: Sr. No. Startup Name Area of Work 1 PrivaSapien Technologies Responsible AI with privacy engineering and privacy-enhancing technologies 2 CoRover.ai Conversational Gen AI platform for contextual, domain-specific human-like responses 3 Staqu Technologies (Jarvis) AI-powered audio-video analytics for surveillance, security, and governance 4 SatSure Analytics Earth Observation using satellite imagery and AI for decision intelligence 5 Storyvord AI-based video content generation from product URLs and brand context 6 VolarAlta Drone-based AI solutions for industrial inspections 7 Smartail AI-powered edtech platform for immersive learning and real-time assessment8 Secure Blink AI-driven cybersecurity with proactive DevSecOps posture management 9 NeuroPixel.AI AI-enabled ultra-fast image editing for fashion and e-commerce 10 Voicing AI Agentic AI platform for enterprise-grade intelligent voice agents *******