Artificial intelligence

Indian Economy glossary

Also called: AI · Topic: Infrastructure: Transport, Communications and Energy · NCERT: Beyond NCERT

Meaning

Artificial intelligence (AI) means computer systems that do tasks which normally need human intelligence. Examples are learning from data, reasoning, understanding and writing language, recognising images and making decisions.

  • Why it matters: governments and firms use AI to work faster, deliver services and make policy.
  • The catch: if no one checks how it is used, AI can also weaken transparency (openness) and accountability (answerability) in government.

Explanation

How it works

  • Learning from data
  • An AI system is trained on very large amounts of data, such as text, images or records.
  • It finds patterns in that data.

  • It then uses those patterns to predict, classify or produce new output.

  • Generative AI is a type of AI that creates new content, such as text, images or summaries. It does not just sort or label existing data.

  • It writes fluent text quickly and in large amounts.

  • That text can sound smooth but say very little.

  • AI follows instructions. It writes whatever it is told to write.

  • Ask it to "highlight achievements, avoid shortcomings" and it gives vague text.
  • Ask it to "state targets and results with numbers" and it gives clear text.

AI as a user of infrastructure

AI sits on top of communication infrastructure (the cables, wireless devices, towers, satellites and data centres that carry messages, images and videos).

  • Data centres store data and run AI models. They were given infrastructure status in 2022, which means they were added to the Harmonised Master List of Infrastructure and can get easier, cheaper long-term loans (verify current).
  • Fibre and mobile networks carry data to and from users. Examples are BharatNet optical fibre to gram panchayats and 5G, launched in October 2022.
  • Electricity powers the servers. So AI depends on the energy infrastructure too.
  • If one link is weak, the whole chain fails. This is the "pieces of a puzzle" idea from Class 7.

AI in government: uses and risks

  • Uses: the G7–OECD position is that AI in the public sector can improve:
  • internal efficiency;
  • how well policy is made;
  • how quickly services respond to citizens;

  • transparency and accountability.

  • Risk 1, "shadow" use: public servants often use generative AI without formal approval or clear oversight. AI-drafted output may then skip the normal checking chain.

  • Risk 2, cheap vagueness: AI makes buzzword-heavy official text cheap and endless. The reader has to dig through more words to find fewer facts.
  • Risk 3, lost trust: the OECD lists public trust, and public resistance to government use of AI, as risks of generative AI experiments.
  • The OECD principle: governments that deploy AI should commit to fairness, accountability and transparency, and keep evaluating the systems after deployment.

In India

  • Policy framework: MeitY released the India AI Governance Guidelines under the IndiaAI Mission. They have seven guiding principles and six pillars of recommendations.
  • The guidelines admit that voluntary frameworks lack legal enforceability. So no ministry is legally bound by them.
  • They suggest tools like transparency reports, audits and self-certification. Sensitive reports would go confidentially to regulators, not to the public.
  • They say AI responsibilities are spread across several agencies, and they propose a whole-of-government approach.

  • It is unclear how liability (legal responsibility for harm) is split between developers, deployers and end-users.

  • No public register: the retrieved Indian sources show no public list of where government uses AI. In contrast:

  • the Netherlands has had a public Algorithm Register since 2021;
  • the UK keeps Algorithmic Transparency Records;

  • the US requires a federal AI use-case inventory, updated at least once a year.

  • The base it builds on: e-governance services such as DigiLocker, the communication network (about 900 million internet connections in 2025), BharatNet and 5G, and the Telecommunications Act 2023.

  • Transparency law alongside it: Section 4 of the RTI Act 2005 requires public authorities to publish key information on their own, without waiting for a request. This is called proactive disclosure, and a third party must audit it every year under a DoPT direction.

Don't confuse with

  • E-governance: this is the government delivering services through communication technology, for example DigiLocker. AI is one tool within e-governance. Not all e-governance uses AI.
  • Generative AI vs AI in general: generative AI creates new content such as text or images. AI in general also includes systems that only classify, predict or recommend.
  • Algorithm register vs AI governance guidelines: a register is a public list of where government uses AI tools (Netherlands, 2021). India's guidelines are a voluntary set of principles, not a public list.
  • Simple automation: automation follows fixed rules written by people, like a FASTag toll gate opening automatically. AI learns patterns from data and can handle tasks it was not given exact rules for.

Prelims Hooks

  • The G7 Toolkit for AI in the Public Sector (2024) was prepared jointly with the OECD. It is not a UN or UNESCO document.
  • The first public national algorithm register named in the sources is the Netherlands' Algorithm Register (2021), not the UK's or the EU's.
  • The US federal AI use-case inventory must be updated at least annually. The UK uses Algorithmic Transparency Records.
  • The OECD's three principles for AI used in public administration are fairness, accountability and transparency.
  • India AI Governance Guidelines (MeitY, IndiaAI Mission): seven principles, six pillars, voluntary and not legally enforceable.
  • Trap: "Data centres are outside the Harmonised Master List of Infrastructure." FALSE. They got infrastructure status in 2022 (verify current).

Mains Points

  • AI is a tool, not the root cause. Evasive official language is "an age-old trick". But AI makes such text cheap to produce, and when no one discloses who used the tool, no one can be held responsible. Possible fixes:
  • a one-line note on each release saying whether AI helped draft it and which officer approved it;
  • a public list of government AI uses, on the lines of the UK and US models;

  • adding press-note quality to the existing RTI Section 4 transparency audit.

  • Voluntary rules or binding law? India's guidelines encourage innovation but cannot be enforced, report mainly to regulators, and leave liability unclear. For GS-II and GS-III answers, weigh light-touch regulation that supports innovation against citizens' right to information.

  • AI needs infrastructure first. AI can improve government efficiency and service delivery, but only on top of data centres, fibre (BharatNet), 5G and reliable power. Without last-mile connectivity, AI-based services could widen the rural-urban digital divide instead of moving towards Kalam's goal of equal infrastructure in urban and rural areas.

Read more