Compare India's principle-based AI Governance Guidelines with the self-regulatory model of the U.S. What does each approach mean for innovation and accountability?

Frontier AI is changing faster than any statute can be written, so governments are trying out soft-law models. On 29 September 2026 the U.S. White House Accord set out a one-page, non-binding pledge by frontier-model developers [1]. India's AI Governance Guidelines (MeitY, IndiaAI Mission) take a principle-based approach that rests on existing law [2]. Both are voluntary, but India's model is better placed to secure accountability.

Comparison of the two models

Feature	U.S. Accord	India's Guidelines
Basis	Corporate pledge of 4 measures [1]	7 Sutras, recommendations across 6 pillars [2]
Scope	Frontier-model builders only [1]	Developers, deployers, regulators [2]
Oversight	Internal team, company-chosen auditor, board committee [1]	Proposed AIGG, TPEC, AI Safety Institute [2]
Legal backing	None; no penalties [1]	Existing laws plus proposed graded liability [2]

Implications for innovation

• U.S.: compliance costs are low and there is no penalty risk, so firms can move fast. Standards will develop as signatories "meet regularly" [1].
• However, standards set by large incumbents can become de facto entry barriers for smaller firms that had no say in them.
• India: the "Innovation over Restraint" Sutra avoids a new AI law that could quickly go out of date [2].
• Its overall spirit, "Do No Harm", aims to keep innovation inclusive and suited to the Global South [2].

Implications for accountability

• U.S.: each company checks itself through people it chooses. Audit results do not have to be published, and liability is never mentioned. This creates a risk of regulatory capture [1].
• India: the guidelines themselves admit that voluntary frameworks lack legal enforceability [2].
• To close that gap, India proposes graded liability based on a firm's role, the level of risk and the care it took. It also proposes public bodies to oversee AI [2].
• Neither model yet makes comparable public disclosure compulsory. The OECD's Hiroshima AI Process Reporting Framework does require it [3], and version 2.0 extends it to SMEs [4].

In short, the U.S. model favours speed at the cost of answerability, while India's model tries to keep both. India should turn graded liability into actual rules and have the AI Safety Institute adopt HAIP-style reporting [3]. It should also certify independent auditors. These steps would combine innovation with the constitutional values of dignity and fairness (Article 21) and support SDG 9 and SDG 16.