India’s AI Governance Shift: Safety, Bias, Accountability

From Promotion to Guardrails: India’s AI Governance Shift

In November 2025, India took a decisive step toward responsible artificial intelligence when the Ministry of Electronics and IT (MeitY) released the India AI Governance Guidelines 2025. This marked the country’s first comprehensive framework focused not just on promoting AI, but on governing it.

By early 2026, these guidelines were formally brought under the IndiaAI Mission, positioning them as a foundational milestone. The approach is explicitly human-centric, anchored in a strong “do no harm” principle. The objective is clear: enable AI-driven economic growth while putting system-level controls in place to address deepfakes, discrimination, and national-security risks.

Safety First: Deepfakes, Misuse, and System Testing

The guidelines clearly identify the biggest AI risks facing India today. These include malicious uses such as deepfakes and misinformation, algorithmic bias, lack of transparency, broader systemic risks, and threats to national security.

Instead of relying on generic global checklists, MeitY calls for an India-specific AI risk-assessment framework, grounded in real evidence of harm. Three safety mechanisms stand out:

• A proposed AI Safety Institute (AISI) focused on testing, red-teaming, and stress-testing AI systems.

• A National AI Incident Database to document and analyse harmful AI incidents

• Techno-legal safeguards, including watermarking, traceability, and mandatory incident reporting

Together, these measures aim to move AI safety from theory to enforcement.

Bias and Inclusion as Non-Negotiables

A central theme of the governance framework is inclusion. The document stresses that India’s social diversity must be reflected in AI systems—and that bias cannot be treated as an afterthought.

The guidelines recommend a risk-based regulatory approach. High-impact sectors such as lending, hiring, healthcare, and policing are expected to meet stricter standards. These include bias testing, use of representative datasets, and independent audits.

Related policy discussions also push for explainable AI, internal ethics boards within organisations, and tighter oversight of AI-enabled surveillance. These ideas align closely with debates around the draft AI (Ethics and Accountability) Bill, 2025, reinforcing consistency across India’s emerging AI policy landscape.

Accountability: Defining Responsibility in AI Systems

One of the most important questions in AI governance is simple but difficult: who is responsible when AI causes harm?

Analyses of the draft AI (Ethics and Accountability) Bill 2025 suggest that India is moving toward clearer accountability. The proposed framework assigns defined responsibilities to both developers and deployers of AI systems. These include preventing bias, ensuring human oversight, and building safeguards against misuse.

The governance structure itself is multi-layered:

• A national-level AI Governance Group for coordination

• Expert committees to handle complex or high-risk use cases

• The AI Safety Institute is to develop standards and testing protocols

Private players would also face obligations to comply with Indian law, publish transparency reports, and maintain grievance-redressal mechanisms.

Balancing Innovation with Regulation

Notably, the India AI Governance Guidelines are framed as a flexible framework, not a rigid law. This gives policymakers room to encourage innovation through sandboxes, public-sector pilots, and research partnerships.

At the same time, the option of a dedicated AI statute remains open if risks escalate. The roadmap also highlights capacity-building—training regulators, judges, and professionals—along with public awareness initiatives, ensuring governance keeps pace with fast-moving technology.

India-AI Impact Summit 2026: Policy in the Open

The India-AI Impact Summit 2026 reflects how central trust and safety have become to India’s AI agenda. Organised around seven thematic “chakras,” the Summit dedicates an entire track to “Safe & Trusted AI.”

This signals a clear shift: safety, bias mitigation, and accountability are no longer side conversations. Working groups under the Summit are expected to feed concrete recommendations back into the IndiaAI Mission, reinforcing a multi-stakeholder approach involving government, industry, academia, and civil society.