How Do PSUs Approach AI Governance Differently from Private Enterprises?

Public Sector Undertakings face AI governance requirements that private enterprises do not. Parliamentary accountability: PSU AI decisions that affect the public can be scrutinised through parliamentary questions and CAG audits. Union consultation: major automation programmes in PSUs typically require consultation with recognised trade unions, adding governance complexity and timeline to AI deployments. Public procurement rules: AI consulting and technology procurement must follow GFR (General Financial Rules) and CVC guidelines, which restrict sole-source procurement and require competitive bidding processes. SEBI oversight for listed PSUs: listed PSUs must disclose material AI investments and risks in annual reports (DPE PSU Guidelines, 2025).

AI consultants working with Indian PSUs must be familiar with these governance requirements. Project design must incorporate public accountability mechanisms, union consultation processes, and transparent procurement. AI systems must be auditable to a standard that can withstand CAG scrutiny. These requirements add complexity but are not negotiable for government clients.

What Are the DPDPA Requirements for Government AI in India?

DPDPA 2023 applies to government entities processing personal data of citizens, with some exceptions for law enforcement and national security functions (MeitY, 2023). For citizen-facing government AI, DPDPA requirements include: legal basis for data processing (typically a statutory function, which provides a valid legal basis without requiring consent); data minimisation in AI systems (using only the citizen data necessary for the AI function); purpose limitation (data collected for one government programme cannot be repurposed for a different AI application without legal basis); and automated decision-making transparency (citizens have a right to know when significant decisions are made by automated AI systems and to seek human review).

The automated decision-making transparency requirement is particularly significant for government AI. AI-generated government decisions, such as tax scrutiny selection, benefit eligibility assessment, or subsidy disbursement, must be explainable to citizens who request an explanation. This requires government AI systems to incorporate explainability mechanisms from the design stage.

[ORIGINAL DATA] In our government AI consulting engagements, the most common design error is failing to include an appeal mechanism for AI-driven decisions. Government AI systems that cannot be challenged by citizens are potentially in violation of natural justice principles, regardless of DPDPA requirements. Building a human review pathway for AI decisions affecting citizens is both a legal requirement and a public trust investment.

What Are MeitY's Cloud Requirements for Government AI?

Government agencies and PSUs must use cloud services from providers empanelled by MeitY under the Government Cloud (GI Cloud) framework. The empanelled providers include NIC cloud (for sensitive and strategic government applications), STQC-certified commercial providers (AWS, Azure, Google Cloud in their India-compliant configurations), and state-specific cloud providers (MeitY GI Cloud, 2025). For AI workloads, this means training and inference infrastructure must be on empanelled cloud providers, and data must remain in India for specified sensitivity classifications. AI consultants serving government and PSU clients must design architectures that comply with these cloud constraints, which may limit access to some AI services available only in non-empanelled regions.

[CHART: Government AI cloud architecture for India - MeitY empanelled providers, NIC cloud, sensitivity classifications, AI service availability - Source: Opsio 2026]

How Should PSUs Structure AI Procurement?

PSU AI procurement must follow GFR 2017 and specific PSU procurement policies. Key principles: competitive bidding for contracts above defined thresholds (typically INR 2 lakh for goods, higher for services); evaluation committees with appropriate technical, financial, and domain expertise; GeM (Government e-Marketplace) preference for standard AI software products; and phased implementation with defined evaluation checkpoints before release of subsequent tranches. PSU AI consulting procurement typically uses an RFP process with quality-and-cost-based scoring (QCBS), where technical quality carries 70-80% weight and price carries 20-30% weight. AI consultants targeting PSU clients must submit competitive bids through the appropriate procurement channel and cannot rely on sole-source awards for significant contracts (DPE, 2025).

<a href="/in/blogs/ai-strategy-roadmap-steps/" title="AI Strategy Roadmap">AI strategy roadmap</a> India

Citation Capsule: AI in Indian Government and PSUs

The INDIAai Mission (INR 10,372 crore) funds compute, datasets, and AI applications for government and national development use cases. Digital India has digitised 1.2 lakh government services, creating data infrastructure for AI. GSTN's AI fraud detection has identified thousands of crore rupees in GST ITC fraud. DPDPA 2023 requires explainability and appeal mechanisms for AI-driven government decisions. PSU AI procurement must follow GFR 2017 and QCBS evaluation under MeitY cloud empanelment constraints (INDIAai, 2024).

Frequently Asked Questions

How do central government ministries access INDIAai Mission resources?

Central government ministries can access INDIAai Mission compute infrastructure through the INDIAai portal for approved AI research and development projects. AI Business Development Centres funded by INDIAai provide subsidised compute and technical support for government AI pilots. The India Datasets Platform provides curated datasets for ministry-specific AI applications. Ministries typically engage with INDIAai through MeitY's Digital India Corporation as the programme management entity (INDIAai, 2024).

Can state governments use AI consulting independently of the central government?

Yes. State governments have independent AI programme authority and budgets. Several states have launched their own AI missions: Karnataka's AI policy, Telangana's T-Hub AI ecosystem, Maharashtra's AI task force, and Tamil Nadu's AI policy are among the most developed. State government AI must still comply with DPDPA (which is central legislation), MeitY cloud empanelment for state government systems using central infrastructure, and Union government data sharing policies for Aadhaar and GSTN data. States can independently procure AI consulting services through their own GFR-equivalent procurement rules.

What are the specific challenges of AI deployment in Indian PSUs?

The four most common challenges are: legacy IT systems (many PSUs run on decades-old ERP, billing, and process control systems that are difficult to integrate with modern AI platforms); union resistance (employee unions may resist AI automation programmes without adequate redeployment and retraining commitments); procurement timeline (GFR-compliant procurement takes 6-18 months for large contracts, delaying AI programme starts); and talent shortage (PSUs struggle to compete with private sector salaries for AI talent, making internal AI capability building difficult).

What makes a government AI project successful in India?

Government AI success factors in India include: clear mandate from senior leadership (Secretary or CMD level sponsorship); defined citizen-facing outcome metrics rather than technology metrics; DPDPA compliance designed in from the start; phased implementation with working system milestones rather than big-bang launches; transparent communication with affected employees and unions; and integration with existing government IT systems (UMANG, DigiLocker, GSTN, ABDM) rather than building parallel infrastructure. Projects that lack senior sponsorship or citizen outcome metrics consistently fail to reach production regardless of technical quality.

Conclusion

India's government and PSU AI journey is uniquely ambitious: scaling AI across the world's most complex governance ecosystem to serve 1.4 billion citizens. The INDIAai Mission provides funding and infrastructure. Digital India provides the data foundation. The challenge is execution within the constraints of government procurement, union consultation, MeitY cloud requirements, and DPDPA compliance.

AI consultants who understand these constraints, not just the technology, are the ones who can help Indian government agencies and PSUs translate AI policy ambition into citizen service reality. The opportunity is enormous. The governance complexity is manageable with the right expertise.

To explore how we structure government and PSU AI consulting engagements, visit our Opsio's AI consulting practice or read our guide on AI Governance for India: DPDPA and EU AI Act.

For hands-on delivery in India, see Opsio ai governance consulting.