AI & Digital Ethics Grade 10 AI Governance Global Frameworks and Indias Policy

Grade 10 | AI & Digital Ethics
Topic: AI Governance: Global Frameworks and India’s Policy

1. The Driving Question

"If AI can decide who gets a loan, which news you see, or even who gets hired—who gets to decide the rules for how it’s built and used? Why does India’s approach look different from Europe’s or the U.S., and what happens when these rules clash?" This isn’t just about laws—it’s about power. Who holds AI accountable when it makes a mistake that hurts real people? And how do countries with different values (like privacy vs. innovation) agree on what’s "fair" for a technology that crosses borders?

2. The Core Idea — Built, Not Listed

Imagine you’re at a street food market in Delhi, where every stall follows its own rules: one checks IDs for chai, another bans phones, and a third lets kids work after school. Now picture an AI system—like a food delivery app—that operates across all these stalls. If the app’s algorithm starts rejecting orders from certain neighborhoods because of biased data, who steps in? The stall owners? The city? The app’s CEO in Silicon Valley?

AI governance is like creating a rulebook for that market—but the market is the whole world. Some countries (like the EU) treat AI like a dangerous machine that needs strict safety checks (their AI Act bans facial recognition in public spaces). Others (like the U.S.) treat it like a startup experiment, letting companies self-regulate unless harm is proven. India’s approach? It’s like a street vendor’s hustle: fast, flexible, and focused on inclusion (e.g., using AI for rural healthcare) but with fewer guardrails. The problem? AI doesn’t respect borders. A biased hiring tool trained in the U.S. could discriminate against Indian job applicants, and India’s laws might not have the teeth to stop it.

Key Vocabulary:
- AI Governance
Definition: The rules, policies, and institutions that guide how AI is developed, deployed, and monitored to protect people and society.
Example: India’s Digital Personal Data Protection Act (2023) requires companies to delete your data if you ask—but doesn’t ban AI from using it in the first place (unlike the EU’s GDPR).
College Shift: In grad school, you’ll debate whether governance should be top-down (laws) or bottom-up (community norms, like Wikipedia’s editing rules).

• Regulatory Sandbox
  Definition: A controlled environment where companies can test AI systems with relaxed rules, but under government supervision.
  Example: The UK’s Financial Conduct Authority lets fintech startups experiment with AI loan approvals—but if the AI discriminates, the company must fix it or shut down.
  College Shift: Sandboxes raise ethical questions: Is it fair to let companies "practice" with real people’s data? Who bears the risk?

• Algorithmic Accountability
  Definition: The principle that AI systems must be explainable, auditable, and liable for harm they cause.
  Example: When Zomato’s AI in India accidentally blocked all Muslim-owned restaurants in 2023, the company had to publicly explain the bug and compensate affected businesses.
  College Shift: In law school, you’ll study cases where courts struggle to assign blame—was it the developer, the company, or the user who misapplied the AI?

• Digital Sovereignty
  Definition: A country’s right to control data, technology, and digital infrastructure within its borders.
  Example: India’s Data Localization Laws require companies like Facebook to store Indian users’ data on servers inside India—so the government can access it if needed (unlike the U.S., where data flows freely).
  College Shift: In international relations, this becomes a geopolitical weapon (e.g., China’s "Great Firewall" vs. the EU’s "Brussels Effect").

3. Assessment Translation

How this appears on tests:
- State Standardized Tests (e.g., Civics/Economics): Short-answer questions comparing two frameworks (e.g., "How does the EU’s AI Act differ from India’s Digital Personal Data Protection Act in regulating facial recognition?"). Look for: - Proficient: Names both laws, cites one specific rule from each (e.g., "EU bans facial recognition in public; India requires consent but doesn’t ban it"), and explains why the difference matters (e.g., "EU prioritizes privacy; India prioritizes innovation").
- Developing: Lists laws but no specifics, or confuses the frameworks (e.g., "Both ban AI").
- Classroom Debates (Common in AI Ethics Units): "Should India adopt the EU’s AI Act? Defend your position with evidence." Teachers look for: - Proficient: Uses two examples (e.g., "India’s rural healthcare AI needs flexibility, but the EU’s rules could prevent bias in hiring tools").
- Developing: Opinion without evidence (e.g., "India should copy the EU because it’s better").
- SAT/ACT (Indirectly): Reading comprehension passages about tech policy, with questions like "The author’s tone in paragraph 3 suggests that India’s approach is…" (Answer: pragmatic but risky).

Model Proficient Response (Short Answer):
Prompt: "Compare how the EU and India regulate AI in hiring tools. Which approach do you think is more effective, and why?" Response: "The EU’s AI Act classifies hiring AI as high-risk, requiring companies to prove their algorithms don’t discriminate before using them. India’s Digital Personal Data Protection Act doesn’t ban biased AI but lets users sue if they’re harmed. The EU’s approach is stricter and prevents harm upfront, but India’s is more flexible for startups. I think India’s approach is better for innovation, but it should add audit requirements like the EU to catch bias early. For example, if a Mumbai company’s AI rejects women for tech jobs, the EU’s rules would stop it before it starts, while India’s might only act after lawsuits."

4. Mistake Taxonomy

Mistake 1: Overgeneralizing Frameworks
Prompt: "Explain one way the U.S. and India’s AI policies differ." Common Wrong Answer: "The U.S. has no rules, but India has strict laws." Why It Loses Credit: - Misreads the question format: The prompt asks for one specific difference, not a vague comparison.
- Factual error: The U.S. has sectoral laws (e.g., FDA rules for AI in medicine) and state laws (e.g., California’s privacy act).
Correct Approach: 1. Pick one policy area (e.g., data privacy).
2. Name one U.S. law (e.g., "California Consumer Privacy Act lets users opt out of data sales") and one Indian law (e.g., "India’s DPDP Act requires companies to delete data on request").
3. Explain the difference: "The U.S. gives users control over data, while India gives them ownership."

Mistake 2: Ignoring Trade-Offs
Prompt: "Why might a country like India resist adopting the EU’s AI Act?" Common Wrong Answer: "Because India doesn’t care about privacy." Why It Loses Credit: - Lacks evidence: Assumes intent without citing India’s actual priorities (e.g., digital inclusion, economic growth).
- No nuance: The EU’s rules could hurt India’s AI startups by adding compliance costs.
Correct Approach: 1. Identify India’s goals (e.g., "India wants to use AI for rural healthcare and financial inclusion").
2. Explain how the EU’s rules conflict (e.g., "The AI Act’s strict testing requirements could slow down Indian startups like Staqu, which uses AI for low-cost medical imaging").
3. Add a counterpoint: "But India might adopt some EU rules, like bans on social scoring, to protect citizens."

Mistake 3: Confusing "Governance" with "Bans"
Prompt: "What is one challenge of global AI governance?" Common Wrong Answer: "Countries can’t agree on whether to ban AI." Why It Loses Credit: - Misunderstands the concept: Governance isn’t about bans—it’s about rules for use (e.g., transparency, accountability).
- No real-world example: Doesn’t cite a specific conflict (e.g., U.S. vs. China on AI in military drones).
Correct Approach: 1. Define governance: "Rules for how AI is developed and used, not just whether it’s allowed." 2. Give an example: "The U.S. and EU disagree on how to regulate AI in hiring—should companies prove their AI is fair before using it (EU), or only after harm is proven (U.S.)?" 3. Explain the challenge: "This makes it hard for global companies like Infosys to follow one set of rules."

5. Connection Layer

1. Within AI & Digital Ethics → AI Bias in Hiring Tools
   Why it connects: India’s policy focuses on inclusion (e.g., using AI to hire from rural areas), but if the AI is trained on biased data (e.g., favoring urban candidates), the policy backfires. Understanding governance helps you spot where bias sneaks in—is it the algorithm, the data, or the lack of rules?

2. Across Subjects → Economics (Globalization)
   Why it connects: AI governance is like trade agreements—countries negotiate rules for a shared resource (data, like oil in the 20th century). India’s "data localization" laws are similar to import tariffs: they protect local industries but can spark trade wars (e.g., the U.S. has threatened sanctions over India’s data rules).

3. Outside School → Social Media Algorithms
   Why it connects: Next time you see a YouTube recommendation or Instagram ad, ask: Who decided this was "relevant" to me? The EU’s AI Act would require platforms to explain why you’re seeing certain content, while India’s rules might only require them to let you opt out. This is governance in action—you’re living inside the rulebook.

6. The Stretch Question

"If India’s AI policy is designed for ‘digital inclusion,’ why do critics say it actually excludes marginalized groups—like low-caste users or rural women—from AI’s benefits?"

Pointer Toward the Answer:
1. Data Gaps: India’s AI systems (e.g., Aadhaar for welfare) rely on government data, but marginalized groups are often missing from these datasets (e.g., rural women without birth certificates). If the AI is trained on incomplete data, it can’t serve them.
2. Bias in Design: India’s policies prioritize scalability (e.g., AI for 1.4 billion people) over fairness audits. For example, facial recognition in police systems has higher error rates for darker-skinned individuals, but India’s rules don’t require testing for this.
3. Power Imbalances: India’s "sandbox" approach lets companies experiment with AI, but marginalized groups lack the legal literacy to challenge harm. For instance, if a loan AI rejects Dalit applicants, they may not know they can sue under the DPDP Act.

This isn’t just about policy—it’s about who gets to shape it. Who sits at the table when India writes its AI rules?