AI Governance

The governance of AI is now a major diplomatic concern. A series of high-level global summits have shaped international norms around AI safety, regulation, and use.

AI Safety Summit 2023 (Bletchley Park, UK)

The AI Safety Summit was the first global-level summit dedicated exclusively to the risks and safety of frontier (advanced) AI systems. It was held at Bletchley Park — historically famous as the World War II code-breaking centre — a symbolically apt location for discussing the codes governing AI.

Why Was It Held?

• Rapid advancement of frontier AI and LLMs raised urgent global concerns about unpredictable behaviour and societal impacts
• Risk of misuse and misinformation: Frontier AI can generate convincing deepfakes, fake news, manipulated content — threatening democracy and public trust
• Cybersecurity and biosecurity threats: Frontier AI may amplify cyberattacks or assist in biological research misuse
• Potential loss of human control: Highly autonomous AI could operate beyond effective human oversight

What is Frontier AI?

📌 Frontier AI — Definition & Features

Frontier AI = the most advanced, cutting-edge AI models with massive scale, high complexity, and general-purpose capabilities. They can learn from vast datasets and perform sophisticated tasks far beyond traditional narrow AI. Key Features: • General-purpose capability — adapts to multiple domains without task-specific retraining • Massive scale — billions of parameters, enormous datasets, advanced semiconductors • Multimodal — processes text, images, audio, sometimes video • Advanced capabilities — strong reasoning, creative generation, zero-shot & few-shot learning • High impact AND high risk — transformative potential, but also deepfakes, cyber threats, accountability gaps

The Bletchley Declaration — World’s First AI Safety Agreement

Feature	Details
Nature	First international political agreement on AI safety — non-binding but norm-setting
Signatories	28 countries + European Union; including USA, China, UK, India, Australia, Brazil, Canada, France, Germany, Japan, Kenya, Saudi Arabia, UAE
Key Recognition	Acknowledges that highly capable frontier AI systems can pose serious societal, economic, and security risks if ungoverned
Core Commitment	Global collaboration among governments, researchers, and industry to manage AI risks that transcend national borders
AI Development Principle	AI should be developed and deployed in a safe, ethical, and human-centric manner with meaningful human oversight
Transparency	Encourages information sharing, scientific research, and evaluation mechanisms to understand frontier AI risks
Inclusivity	Recognises the role of governments, academia, and private AI developers — not just governments

AI Seoul Summit 2024 (South Korea)

The AI Seoul Summit was the second global high-level meeting on AI, co-hosted by South Korea and the UK. It expanded the global AI agenda beyond safety to include innovation and inclusivity.

• Seoul Declaration: Signed by 10 nations + EU — affirmed commitment to safe, innovative, and inclusive AI; emphasised interoperability between governance frameworks
• Historic First: 16 global tech firms (Google, Meta, xAI, Microsoft, OpenAI, Samsung, etc.) voluntarily committed to ‘kill switches’ — stopping AI development/deployment if risks cross ‘intolerable’ thresholds

Global Partnership on Artificial Intelligence (GPAI)

GPAI is a multi-stakeholder international initiative aiming to bridge the gap between theory and practice on AI by supporting research and applied activities.

Aspect	Details
Launch	Officially launched in 2020 with 15 founding members including India, US, and UK
Current Members	~29 international partners
OECD Partnership	In 2024, GPAI and OECD announced integrated partnership to advance trustworthy AI globally
GPAI Plenary	Primary working body — all members; meets twice a year; discusses AI governance, endorses policy, decides new membership
Annual Summit	One major summit per year — global platform for sharing progress and brainstorming projects

Key Focus Areas of GPAI

• Responsible AI: Ethical, transparent, accountable, human-centric AI aligned with human rights and democratic values
• Data Governance: Data quality, access, privacy, sharing, and cross-border data issues
• AI & Future of Work: Impact on employment, skills, labour markets, workforce transitions, reskilling
• Innovation & Commercialisation: AI research, startups, industry adoption — balancing innovation with responsibility
• AI for Social Good: Healthcare, agriculture, education, climate action, inclusion

4th GPAI Annual Summit 2023 — New Delhi Declaration

🇮🇳 India’s Global AI Leadership Moment

India hosted the 4th GPAI Annual Summit 2023 at Bharat Mandapam, New Delhi — the FIRST GPAI summit held in the Global South. It concluded with the unanimous adoption of the New Delhi Declaration. Key Features of New Delhi Declaration: • Safe, Secure & Trustworthy AI — risk mitigation, safety testing, responsible deployment • Balanced Approach (Middle Path) — unlike Bletchley’s safety-centric focus, New Delhi advocated balancing safety with innovation and development benefits • AI for Development — healthcare, agriculture, education, public services • Agriculture as New Priority — formally added as GPAI thematic focus • Inclusive & Global South-Centric — equitable AI access, capacity building for LMICs • Non-binding nature — political commitment, not legally binding treaty

AI Action Summit 2025 (Paris, France)

The AI Action Summit was held at the Grand Palais, Paris in 2025. France and India co-chaired the summit. It marked a decisive shift in the global AI governance conversation.

Key Outcome	Details
Strategic Shift	From ‘AI Safety’ → ‘AI Action’: moved from risk-centric debates to implementation-focused governance emphasising practical deployment and real-world impact
First International AI Safety Report	First comprehensive, independent scientific report on AI capabilities and risks — published and discussed at this summit
Inclusive & Sustainable AI Declaration	Signed by ~58 countries including India and China. The US and UK did NOT sign — citing concerns about over-regulation and national security
Current AI Initiative	Aims to develop open, ethically governed AI models for public interest areas like healthcare and education
Coalition for Sustainable AI	Led by France, UNEP, and ITU — addresses AI’s significant energy consumption and environmental impact
Global Framework for AI Environmental Impact	First time environmental impact became a core AI governance theme — measuring energy consumption, promoting green AI, sustainable data centres

AI for Good Global Summit (Annual, Geneva)

AI for Good Global Summit is a flagship UN-led global platform promoting AI for advancing the Sustainable Development Goals (SDGs).

Organised by the International Telecommunication Union (ITU) in partnership with the United Nations, it has been held annually in Geneva since 2017.

• AI Governance Day: A dedicated high-level policy forum within the summit — focuses on governance, standards, ethics, and regulatory coordination for AI
• Key Objective 1: Harness AI for sustainable development — health, education, climate, disaster management, poverty, inclusion
• Key Objective 2: Bridge the gap between innovation and impact by connecting AI innovators, policymakers, UN agencies, academia, and civil society
• Key Objective 3: Promote ethical and inclusive AI — especially for developing countries in the Global South
• Key Objective 4: Capacity building — skills development, knowledge sharing, and access to AI technologies for developing nations

Global IndiaAI Summit 2024

Organised by MeitY under the IndiaAI Mission in 2024, coinciding with India’s role as Lead Chair of GPAI for that year.

• Positioned India as a global AI leader — a hub for responsible, development-oriented, and human-centric AI
• Facilitated collaboration among governments, industry, startups, academia, and international organisations
• Highlighted AI use-cases for public good: healthcare, agriculture, education, governance, climate
• Encouraged innovation in Indian startups, research institutions, and domestic AI capabilities

India-AI Impact Summit 2026

The India AI Impact Summit (Feb 16–21, 2026, New Delhi) positioned India as a leader in global AI governance, shifting focus from a primarily Western, risk-centric dialogue to one centered on inclusivity, development, and the Global South.

India’s AI Vision & Priorities

The summit was structured around 3 core Sutras and 7 Chakras (pillars of action). India articulated four integrated priorities:

• Democratising AI: Preventing the concentration of AI capabilities in a few tech monopolies.
• Global South Inclusion: Building capabilities for under-represented languages in Large Language Models (LLMs).
• Safe and Trusted AI: Establishing ethical governance frameworks.
• Sovereign Capacity Building: Developing indigenous LLMs, data centers, domestic semiconductor ecosystems, and compute infrastructure.

3 Sutras and 7 chakras

• The Three Sutras (Pillars): People, Planet, and Progress.
  • People: Developing AI that empowers citizens through healthcare, education, and financial inclusion. 
  • Planet: Using AI for sustainable practices, climate resilience, and resource efficiency. 
  • Progress: Harnessing AI for economic growth, governance, and public service delivery. 
• The Seven Chakras (Working Groups): Focused themes including Health, Agriculture, Safe & Trusted AI, Science, Inclusion, Democratizing AI Resources, and Economic Development. 

Major Outcomes & Strategic Shifts

1. The New Delhi Declaration on AI Impact

A landmark diplomatic outcome endorsed by 89 countries and international organizations.

• Core Philosophy: Rooted in “Sarvajan Hitaya, Sarvajan Sukhaya” (Welfare for all) and “Vasudhaiva Kutumbakam” (The world is one family).
• Nature: All commitments are voluntary and non-binding to encourage participation from developing nations.
• Key Frameworks Established:
  • Charter for the Democratic Diffusion of AI: Ensures equitable access to foundational AI resources globally.
  • Global AI Impact Commons & Trusted AI Commons: Platforms to share AI use cases, benchmarks, and best practices.
  • International Network of AI for Science Institutions: Accelerates AI-driven scientific collaboration.
  • AI Workforce Development Playbook: Focuses on reskilling workers for the AI economy.
  • Guiding Principles on Resilient and Efficient AI: Promotes sustainable and energy-efficient AI infrastructure.

2. Geopolitical Assertion: The Pax Silica Initiative

• India formally joined the Pax Silica Initiative, a U.S.-led framework designed to secure supply chains for semiconductors, advanced computing hardware, and critical minerals.
• Significance: Marks a major geopolitical shift in response to technological decoupling. It aligns India with trusted technology coalitions to reduce overdependence on concentrated manufacturing hubs (like China).
• Bilateral Tie-in: The summit also saw the announcement of the India-U.S. AI Opportunity Partnership for regulatory cooperation, R&D, and data center collaboration.

3. Economic and Technological Milestones

• Massive Capital Influx: Secured ~$250 billion in AI infrastructure pledges and $20 billion in deep-tech venture funding.
• Corporate Partnerships:
  • Domestic: Reliance and Adani committed to massive AI infrastructure.
  • Global: OpenAI partnered with Tata Group (data centers); Anthropic partnered with Infosys.
• Technological Sovereignty (Sarvam AI): Backed by the IndiaAI Mission, Sarvam AI launched indigenous open-source LLMs, demonstrating frugal innovation and marking India’s shift from an AI consumer to a foundational model creator.

Analytical Takeaway

The 2026 Summit underscores India’s dual strategy in the digital age: aggressively building sovereign AI capabilities (domestic data centers, indigenous LLMs) while simultaneously positioning itself as the diplomatic voice of the Global South (ensuring AI governance is not monopolized by the West). Joining Pax Silica officially links India’s AI ambitions to its broader geostrategic alignment with the US on critical and emerging technologies (iCET).