GS Paper 2
Syllabus: Governance
Context: The Economic Advisory Council to the Prime Minister (EAC-PM) has released a paper “A Complex Adaptive System Framework to Regulate AI”. Also, the European Union (EU) has reached a historic provisional deal on the world’s first comprehensive laws to regulate artificial intelligence (AI).
EAC-PM paper proposes a novel AI regulation framework based on a Complex Adaptive System, with five key principles:
- Specialist Regulator: Establishing an expert regulatory body with a broad mandate and swift responsiveness.
- Transparency: Open licensing of core algorithms for external audits and continuous monitoring of AI systems to ensure transparency.
- AI Accountability: Mandating standardized incident reporting protocols to document system aberrations or failures and define clear lines of accountability.
- Manual ‘Overrides’ and ‘Authorization Chokepoints’: Empowering humans to intervene when AI systems behave erratically through mandated manual controls.
- Guardrails and Partitions: Implementing clear boundary conditions to limit undesirable AI behaviours.
Key Components of the EU Framework for AI Regulation:
| Components | Details |
| Safeguards in Legislation | Empowerment of Consumers: Individuals have the ability to launch complaints against perceived AI violations. |
| Clear boundaries are set on AI usage by law enforcement agencies. | |
| Strong restrictions on facial recognition technology and AI manipulation of human behaviour are specified. | |
| Tough penalties for companies found breaking the rules. | |
| Limitations on governments for use of real-time biometric surveillance in public areas (allowed only in cases of serious threats like terrorist attacks) | |
| Categorization of AI Applications | AI applications are classified into four risk categories based on their level of risk and invasiveness. |
| 1. Banned Applications: Mass-scale facial recognition and behavioural control AI applications are largely banned, with exemptions for law enforcement. | |
| 2. High-Risk Applications: Allowed with certification and provision for backend technique transparency, such as AI tools for self-driving cars. | |
| 3. Medium-Risk Applications: Deployable without restrictions, like generative AI chatbots, with explicit disclosure to users about AI interaction, transparency obligations, and detailed tech documentation. | |
| 4. Low-risk applications with transparency obligations. | |
| Other EU’s Regulatory Achievements | General Data Protection Regulation (GDPR) Implementation: Enforced since May 2018, focusing on privacy and explicit consent for data processing. |
| Digital Services Act (DSA): Focus on regulating hate speech, counterfeit goods, etc. | |
| Digital Markets Act (DMA): Identifying “dominant gatekeeper” platforms and addressing non-competitive practices and dominance abuse. |
Concerns related to AI use:
| Concerns | Explanation |
| Privacy | Risk of personal and sensitive information being used unethically, such as for targeted advertising or political manipulation. |
| Responsibility | AI models generating new content may lead to challenges in identifying responsibility, causing ethical dilemmas over the content’s origin. |
| Automation and Job Displacement | AI’s potential to automate processes raises concerns about job displacement, impacting those skilled in the automated areas. |
| Bias and Discrimination | AI training on biased data may result in decisions that unfairly disadvantage certain groups, perpetuating societal inequalities and discrimination. |
| Lack of Transparency | Concerns about accountability for AI actions, questioning whether responsibility lies with creators, deploying companies, or regulating governments. |
India’s Stance on AI Regulation:
India aims to develop sovereign AI and AI computing infrastructure, focusing on real-life applications in healthcare, agriculture, governance, language translation, etc., to catalyse economic development.
Currently, India is transitioning from a no-regulation stance on AI to actively formulating regulations based on a risk-based, user-harm approach. India’s AI penetration factor is 3.09, the highest among G20 and OECD countries. Some of the initiatives are:
- Awareness Initiatives like #AIFORALL
- NITI Aayog’s National Strategy for AI (2018) focused on inclusive and responsible AI
- NITI Aayog’s ‘Principle of Responsible AI,‘ outlines seven key principles for AI use
- Meity launched IndiaAI in 2023, covering all AI-related research.
- TRAI proposed a domestic statutory authority for AI regulation in July 2023, emphasizing a risk-based framework.
- Ethical guidelines for AI in healthcare by the Indian Council of Medical Research
- SEBI’s circular guiding AI policies in the capital market
- National Education Policy 2020 recommends integrating AI awareness into school courses.
Initiatives by other countries:
| Country | Initiative |
| G7 | The Group of Seven (G7) has proposed a “risk-based” regulation for artificial intelligence (AI) tools, which could be a first step towards creating a template to regulate AI such as OpenAI’s ChatGPT and Google’s Bard |
| Italy | Became the first major Western country to ban OpenAI’s ChatGPT due to privacy concerns. |
| UK | Adopts a ‘light-touch’ approach to encourage innovation in the AI industry. |
| Japan | Takes an accommodative approach to AI developers. |
| China | Drafted a 20-point plan to regulate generative AI services, expected to be enforced later this year. |
| US | Proposed a nonbinding roadmap, the AI Bill of Rights Blueprint, outlining five core principles for the responsible use of AI. |
Conclusion:
Although the risks of AI are widely known, it remains unclear how the AI regulations would address these risks and how grievances would be remedied. Previously, Elon Musk, Steve Wozniak, and over 15,000 others had called for a six-month pause in AI development, and for shared safety protocols to be implemented by labs and independent experts.
Insta Links:
Prelims Links:
Q1. With the present state of development, Artificial Intelligence can effectively do which of the following? (UPSC 2020)
- Bring down electricity consumption in industrial units
- Create meaningful short stories and songs
- Disease diagnosis
- Text-to-Speech Conversion
- Wireless transmission of electrical energy
Select the correct answer using the code given below:
(a) 1, 2, 3 and 5 only
(b) 1, 3 and 4 only
(c) 2, 4 and 5 only
(d) 1, 2, 3, 4 and 5
Ans: B
