Key Facts and Data Points

  • Sarvam Vision: 3 billion‑parameter vision‑language model; supports all 22 official Indian languages.
  • Benchmark Scores: 84.3% on olmOCR‑Bench (beats Google Gemini 3 Pro), 93.28% on OmniDocBench v1.5.
  • Bulbul V3: Advanced Text‑to‑Speech model; >35 professional voices across 11 Indian languages, target to cover all 22 scheduled languages.
  • IndiaAI Mission: ₹10,300 crore programme to build a 70‑billion‑parameter LLM (Sarvam‑Large, Sarvam‑Small, Sarvam‑Edge) for population‑scale deployment.
  • Sovereign AI Definition: Ability of a nation to develop, deploy and regulate AI using domestic data, talent, hardware and legal frameworks.

Background and Context

  • The Government of India selected Sarvam AI to develop the first indigenous LLM under the IndiaAI Mission.
  • Global AI models (Google Gemini, OpenAI’s ChatGPT) exhibit Western hallucinations and rely on foreign data centres, raising concerns over data security and cultural relevance.
  • India’s AI strategy aligns with broader initiatives: India Semiconductor Mission, Digital Personal Data Protection (DPDP) Act, 2023, and leadership in the Global Partnership on Artificial Intelligence (GPAI).

Significance for India / Governance / Policy

  • Strategic Autonomy: Reduces dependence on US/China AI giants, safeguarding critical data (e.g., Aadhaar, financial records).
  • Frugal Innovation: High performance with a 3 B‑parameter model demonstrates cost‑effective AI suitable for a developing economy.
  • Digital Inclusion: Multilingual models and region‑sensitive TTS bridge the digital divide, enabling voice‑based access for non‑English speakers.
  • Economic Impact: Indigenous AI can spur a domestic AI industry, create jobs, and attract patient capital for deep‑tech R&D.
  • Policy Measures Needed:
  • Link AI development with the India Semiconductor Mission for domestic AI chips.
  • Strengthen data residency through strict enforcement of the DPDP Act.
  • Establish an AI Safety Institute for bias and safety certification.
  • Mandate use of indigenous AI solutions in public procurement (railways, defence, postal services).

Related Constitutional / Legal Provisions

  • Article 19(1)(a) – Freedom of speech and expression; AI‑driven content platforms must respect this while ensuring no hate speech.
  • Digital Personal Data Protection Act, 2023 – Provides legal basis for data localisation, supporting sovereign AI.
  • National Security Act & IT Act, 2000 – Can be invoked for AI systems handling critical infrastructure.

Conclusion

Sovereign AI is the digital counterpart of national defence, essential for India to transition from a data supplier to a creator of indigenous intelligence. The Sarvam AI breakthroughs illustrate how strategic policy, frugal innovation, and robust governance can together achieve true Atmanirbhar AI.

Drishti Mains Question: "Sovereign AI is the digital equivalent of national defence in the 21st century." Discuss this statement in light of recent developments in indigenous AI models.