Shaping the Future of AI Governance

China has proposed establishing WAICO, a World Artificial Intelligence Cooperation Organization headquartered in Shanghai, positioning itself as the architect of global AI governance rules. This initiative, announced at the APEC Summit, is Beijing's latest bid to rewrite multilateralism in its favor, following a series of China-led global frameworks. While promising technology-sharing and funding for the Global South, the proposal raises critical questions about transparency, control, and whether it will complement or compete with UN-led AI governance efforts. For India, the challenge is clear: engage without endorsing, demand transparency over geography, and ensure access is not contingent on allegiance. As AI becomes the currency of geopolitical influence, the rules being written today will determine who shapes innovation tomorrow and India must help write them or risk living under rules others author. 

What are the Existing Global Governance Mechanisms for Artificial Intelligence? 

• International Organizations-Soft Law & Principles: Global institutions are developing ethical frameworks and non-binding principles to guide national AI governance, promoting human-centric development and shared global values.  
  • This soft law approach encourages consensus but lacks enforcement mechanisms, slowing progress toward concrete accountability. 
  • The UNESCO Recommendation on the Ethics of AI (2021) was the first global standard, adopted by 194 Member States.  
  • Similarly, in 2024 the UN General Assembly adopted a resolution on the promotion of “safe, secure and trustworthy” artificial intelligence (AI) systems that will also benefit sustainable development for all. 
• Regional Hard Law-The EU AI Act Model: The European Union’s AI Act is the world’s first comprehensive, risk-based legal framework for AI, classifying systems by potential harm to determine regulatory obligations.  
  • This hard law framework creates a Brussels Effect, setting a de facto global benchmark, though critics argue it may stifle innovation in certain sectors. 
  • The Act will be fully applicable by June 2026, with bans on “unacceptable-risk” systems (like social scoring) coming into force earlier. 
    • Generative AI models (such as GPT-4) that pose systemic risks must comply with evaluation and reporting norms, especially those trained on more than 10²⁵ floating-point operations (FLOPs). 
• Global Summits & Safety Initiatives: High-level international summits are driving voluntary collaboration between governments and AI developers to address frontier risks.  
  • While these initiatives encourage rapid risk mitigation, they often lack Global South participation and legally binding commitments. 
  • The Bletchley Declaration (UK, 2023) and AI Seoul Summit (2024) brought together major powers to cooperate on AI safety research.  
    • The upcoming India–AI Impact Summit (2026) will be the first large-scale AI summit in the Global South, aiming for a more inclusive, sustainable, and human-centric governance agenda. 
• Multilateral Groupings-Geopolitical Influence: Forums such as the G7 and BRICS are leveraging AI governance to project geopolitical and economic influence, creating parallel governance pathways.  
  • However, this growing fragmentation risks a “splinternet” of AI regulations, complicating global interoperability and trade. 
  • The G7 Hiroshima AI Process (2023) launched an International Code of Conduct for AI developers, emphasizing trustworthy AI aligned with democratic values.  
  • Meanwhile, BRICS, under Brazil’s 2025 presidency, is championing South–South cooperation and BRICS Leaders' Declaration on Global Governance of Artificial Intelligence has gained prominence. 
• Industry Self-Governance & Standards: The private sector plays a vital (though sometimes self-interested) role in setting technical standards and ethical commitments. 
  • While this ensures agility in managing emerging risks, it raises concerns over regulatory capture and limited public accountability. 
  • Major tech firms have created mechanisms like the Frontier AI Safety Commitments, pledging responsible testing and deployment.  
  • Additionally, the US Executive Order 14110 (2023) requires developers of powerful AI models to notify the government, marking a shift toward greater oversight and accountability in high-risk AI systems. 

What are the Key Issues Hindering Unified Global Framework for AI Governance?  

• Geopolitical and Ideological Fragmentation: The major global powers have fundamentally divergent values and strategic objectives for AI, which prevents consensus on core regulatory principles.  
  • The competition for AI supremacy between the US and China forces other nations to align, causing regulatory divergence in areas like data privacy and military applications.  
  • The EU's AI Act emphasizes fundamental rights with fines up to 7% of global revenue, setting a cautious precedent.  
    • Conversely, the US Executive Order 14110 prioritizes national security and innovation, creating an ideological divide on key issues such as cross-border data flows and state-access mandates. 
• Pace of Technological Advancement vs. Law: The exponential speed of AI innovation, particularly in generative models, consistently outpaces the slow, deliberative legislative process, making any unified governance framework quickly obsolete.  
  • By the time international treaties or comprehensive laws are finalized, the underlying technology has already advanced to present new, unforeseen risks. 
  • Model capabilities, such as those powering GPT-5 or newer systems, are estimated to increase dramatically, while the EU AI Act took over three years to pass.  
    • The cost to train frontier AI models is reported to be doubling every nine months, outpacing regulatory adaptation speed and necessitating agile, outcome-based rules. 
• Resource and Infrastructure Disparities: A deep economic and infrastructural divide exists where most AI development, compute power, and talent are concentrated in the Global North, leaving the Global South to bear the risks of deployed, un-audited systems.  
  • This disparity leads to governance frameworks that do not address the developmental priorities or capacity constraints of low-income nations.  
  • IMF analysis suggests AI's economic dividends could be more than twice as large in advanced economies compared to low-income countries, demonstrating the widening gap.  
• Socio-Cultural Dimensions of AI Bias: There is no global consensus on defining core ethical concepts like "fairness," "bias," or "explainability" in AI, as these terms are deeply rooted in diverse cultural and legal traditions.  
  • Moreover, the absence of universal norms also gives rise to social challenges — including algorithmic discrimination in hiring, gender and caste bias in automated decision-making, exclusion of marginalized communities from digital benefits, and reinforcement of stereotypes through biased datasets.  
  • For instance, an AI trained primarily on US or European data often exhibits performance degradation and bias when deployed in African or Asian contexts due to lack of representative data. 
• Data Sovereignty and Cross-Border Flow Conflicts: National laws asserting data sovereignty and privacy rights impose conflicting rules on cross-border data flows, which are the lifeblood of large-scale AI model training.  
  • This regulatory Balkanization forces AI companies to create regional silos, fragmenting the global AI ecosystem instead of unifying it.  
  • The EU's GDPR and the new US regulations targeting data transfers to "countries of concern" (like China) exemplify this trend.  
  • South Korea's Personal Information Protection Commission has previously ordered an international fintech company to destroy AI models trained with personal data acquired improperly, demonstrating the legal risks of cross-border data use. 
• Technical Opacity and Accountability Bottlenecks: The "black box" nature of complex foundation models hinders the ability of regulators to effectively audit, explain, or assign liability for autonomous AI decisions, undermining traditional legal mechanisms.  
  • Without standardized technical requirements for transparency and traceability, enforcement of any global rule remains tenuous.  
  • The latest Large Language Models (LLMs) can have trillions of parameters, making their decision-making processes virtually opaque to human inspectors. 
    • This opacity complicates compliance with "explainability" and "human oversight" principles found in regulations like the EU AI Act, shifting the burden of proof away from the developer.

What Measures can be Adopted to Progress Towards a Responsible and Inclusive Global AI Governance Framework? 

• Establishing a Polycentric 'Governance Commons': The core argument is that attempting a single, monolithic global regulator is infeasible due to geopolitical resistance; thus, the solution lies in building a polycentric governance commons.  
  • This involves creating interoperable, specialized institutions that focus on narrow, high-impact areas like safety or standards, allowing global cooperation without sacrificing national sovereignty.  
  • This decentralized, 'networked' model facilitates agile coordination between existing bodies (UN, G7, OECD, etc.) to address systemic risks while respecting diverse regulatory approaches. 
• Risk-Adaptive Regulatory Sandboxes (ARRS): A critical measure is the establishment of Risk-Adaptive Regulatory Sandboxes (ARRS), distinct from simple Innovation Sandboxes.  
  • Governance must be a living, iterative process that constantly recalibrates rules based on real-world system performance and potential harm escalation.  
  • These sandboxes allow high-risk models to be tested in controlled, simulated environments, with rules automatically tightening or relaxing based on demonstrable safety metrics and compliance adherence, ensuring regulation keeps pace with technological speed. 
• Global South-Led AI Capacity Hubs: To ensure inclusivity, a Global South-Led AI Capacity Hub measure must be adopted.  
  • Mere inclusion in discussions is insufficient; true inclusion requires sovereign technical and regulatory capacity.  
  • These hubs, financially backed by developed nations and international finance institutions, would focus on developing open-source, decolonized Foundation Models and training regulators from developing nations on data governance, model auditing, and localized risk assessment. 
• Mandatory Digital Provenance and Labeling: A key practical measure is mandating Digital Provenance and Labeling Standards for all AI-generated content and high-risk models.  
  • Trust is operationalized through traceability; users and regulators must be able to verify the origin and development history of AI outputs.  
  • This involves creating a global, standardized "AI Passport" or metadata system to verify whether content is synthetic (deepfakes) and document a model’s training data and bias audits, making accountability a technical requirement. 
• 'Techno-Legal' Compliance Integration: Countries must implement 'Techno-Legal' Compliance Integration, which embeds governance directly into the AI development lifecycle (MLOps).  
  • This involves creating and sharing standardized Policy-as-Code libraries and automated compliance tools that check for ethical criteria, bias, and legal requirements (like data privacy), automatically flagging non-compliance during model development, making responsible AI the default setting. 
• Standardized Model Card and Audit APIs (Application Programming Interfaces): A crucial technical measure is to globally mandate Standardized Model Card and Audit APIs for all foundation models.  
  • Technical compliance must be machine-readable and universally comparable, moving beyond opaque, document-based disclosures.  
  • These standardized APIs would allow independent third-party auditors and regulators across the globe to programmatically access, test, and verify key governance parameters, such as bias metrics, safety guardrails, and provenance data, fostering a culture of verifiable transparency. 

Conclusion:

AI governance is fast emerging as the new arena of global power and principle. As nations race to shape its rules, India must champion an inclusive, human-centric, and equitable framework that bridges the Global North–South divide. By aligning innovation with ethics and sovereignty with cooperation, India can help script a shared AI future. The moment to act is now, before the code of tomorrow’s world is written by a few. 

Drishti Mains Question: “As Artificial Intelligence becomes a new arena of global power, shaping its governance framework is as much about ethics as it is about geopolitics.” Discuss India’s role and strategy in ensuring an inclusive and equitable global AI order.