Wednesday, 18 November 2026 | Asia's First Monthly Magazine on e-Governance · Est. 2005
SAS Digital Sovereignty
Gujarat AI Summit
SAS Digital Sovereignty
Gujarat AI Summit
Live — 2,069 transfers tracked Updated 18-11-2026, 14:46
Live
Deputation J&K transfers JKAS officer Sabha Sarwar on deputation to Ladakh Today Transfer Maharashtra reshuffles three senior IAS officers across Mantralaya departments Today Transfer Uttarakhand reshuffles six IAS officers, Nitesh Kumar Jha named Secretary Urban Development Today Appointment Malay Jain Named Balaghat SP in Madhya Pradesh IPS 5d ago Appointment Ajay Kumar Yadav Appointed CRPF Additional Director General 9d ago Promotion Punjab Elevates Four Senior IPS Officers to DGP Rank 10d ago Appointment K. Lakshmipriya Named GBA Special Commissioner, Karnataka 11d ago Transfer Bihar Home Department transfers Khagaria SP Bhanu Pratap Singh to Police Headquarters 13d ago Transfer Jharkhand transfers seven IAS officers, Anup Garg named Joint Secretary in Building Construction 13d ago Appointment Girish S.N. Named Odisha Health Commissioner-Secretary 13d ago Extension Maharashtra reshuffles 16 police officers, three Thane seniors get one-year extension 13d ago Posting Shailendra Kumar gets interim charge of J&K Chief Secretary post 13d ago Transfer Uttar Pradesh transfers two IPS officers across Noida and Kanpur Nagar commissionerates 14d ago Resignation Madhya Pradesh IPS officer Anurag Sujania resigns four days after transfer from Sagar 14d ago Transfer Tamil Nadu transfers three central prison superintendents amid escape incidents 14d ago Transfer Bihar Reshuffles Four Senior IAS Officers Across Key Departments 14d ago Posting 1991-Batch IAS Officer CK Anil Takes Charge as Bihar Education ACS 14d ago Transfer Odisha IPS Reshuffle: Jail DG Sushant Nath Moved to Home Department as OSD 14d ago Transfer J&K Reshuffles 12 IAS, IPS and JKAS Officers Across Key Departments 15d ago Transfer UP IAS Reshuffle: Alok Kumar-II Becomes IIDC, 15 Officers Moved 15d ago Transfer Kerala Reshuffles 35 IAS Officers, Jyothilal Gets Home Department 15d ago Transfer Karnataka Transfers 9 Officers; Latha R. Appointed Housing Board Commissioner 29d ago Transfer UP Transfers 7 IAS Officers; Dhirendra Singh Sachan Gets Panchayati Raj Charge 30d ago Transfer Delhi Assigns New Roles to 3 IAS Officers; Nandini Paliwal Gets Power Department 30d ago Transfer MP Transfers 4 IAS Officers; Ramprakash Ahirwar Gets Bhopal Municipal Commissioner Post 30d ago Transfer Andhra Pradesh Transfers 11 IAS Officers; B. Anil Kumar Reddy Gets Industries Post 13 Oct Appointment Centre Approves 26 Senior Appointments; Manisha Trighatia Gets Urban Development Role 12 Oct Transfer J&K Transfers 49 IAS, JKAS Officers; Mamta Yadav Gets Information Department 12 Oct Transfer Chhattisgarh Transfers 18 IPS Officers; Jha Amankumar Ramankumar Gets SIA Charge 10 Oct Transfer Delhi Transfers 34 IAS, DANICS Officers; Awanish Kumar Gets Excise Charge 9 Oct

Hover a headline to pause it · drag the rail to scrub · ← → to step · Space to pause

From Oversight to Insight: Rethinking Responsible AI Governance

An artificial intelligence system may meet every compliance requirement when it is launched, but that does not guarantee that it will remain safe, fair or reliable over time. Its data can change, users may interact with it differently, and the environment in which it operates can evolve. In some cases, the consequences of its decisions may change faster than existing governance processes can respond.


This raises an important question for organisations adopting AI: is it enough to establish rules, conduct audits and check whether systems meet prescribed standards, or should governance also provide the ability to identify risks and act before they cause harm?

As AI moves into areas such as banking, recruitment, healthcare and public service delivery, the limitations of traditional compliance-led approaches are becoming more visible. Responsible AI now requires a shift from simply checking whether rules are being followed to understanding whether those rules are actually protecting people and institutions.


Also Read: From AI Research to Real-World Impact: Opportunities for India’s Research Ecosystem


Why Traditional Governance Needs to Evolve

For a long time, technology governance has largely relied on oversight mechanisms such as policies, controls, audits and compliance checks. These remain essential, but they cannot independently address the risks associated with systems that continue to learn, change and interact with different environments.

Consider an AI system used by a bank to assess creditworthiness, a recruitment platform screening thousands of applications or a government department prioritising access to public services. A system may perform well during its initial assessment, but its outcomes could change as the underlying data or user behaviour evolves.


It may begin producing biased results for a particular community, experience a decline in performance or make decisions that were not anticipated during development. With increasingly autonomous AI agents, the concern extends beyond incorrect answers to actions taken without appropriate authorisation.

Discovering such problems only during a scheduled audit means that governance has documented a failure rather than prevented it.

The distinction between oversight and insight is therefore important. Oversight establishes whether an organisation is following its rules, while insight helps determine whether its systems are functioning as intended and whether corrective action is needed.

The Growing Gap Between AI Adoption and Accountability

The expansion of AI adoption has made this challenge more urgent. As discussed during the address, the Stanford AI Index 2026 recorded an increase in documented AI incidents from 233 in 2024 to 362 in 2025, representing a rise of approximately 55 per cent.

The OECD’s 2026 Digital Government Outlook was also cited to show that 35 of 36 OECD countries were using AI in at least one government process. However, transparency mechanisms and formal registers documenting government AI use remain underdeveloped in several countries.

These developments point to a gap between adopting AI and maintaining visibility over how it is being used. Knowing that an organisation or government department uses AI is not the same as knowing which systems are operating, what decisions they influence, who is responsible for those decisions and whether the outcomes are improving.

Reliability also depends on context. A model that performs well in a demonstration may not be sufficiently reliable for a medical, financial or public-sector decision. The level of scrutiny must therefore reflect the purpose and potential impact of each application.

From Periodic Assessments to Continuous Assurance

Responsible AI governance needs to evolve across three important areas.

The first is the move from periodic assessments to continuous assurance. A system cannot be considered safe simply because it performed well when it was launched. Its performance, security and outcomes need to be monitored as data, users and operating conditions change.

The second is the shift from principle-based commitments to evidence-based decisions. Most organisations recognise the importance of fairness, transparency and accountability. However, these principles need to be supported by measurable indicators that demonstrate whether they are being followed in practice. Organisations must be able to show evidence of fair outcomes, acceptable error rates and appropriate safeguards.

The third is moving governance from a gatekeeping function to an enabling role. Responsible AI should not enter the development process only when a product is nearly ready for the market. Governance needs to be considered from the beginning, including product design, procurement, deployment and the measurement of outcomes.

This approach allows organisations to identify risks early and build safeguards into their systems rather than treating compliance as a final approval exercise.

A Practical Framework: Map, Measure, Monitor, Intervene and Learn

A useful approach to insight-driven governance can be built around five connected steps: Map, Measure, Monitor, Intervene and Learn. Together, these steps create a continuous process through which organisations can understand their AI systems, assess their risks and respond when something goes wrong.

  • Map: Establish visibility

Organisations first need to know where AI is being used. This requires an inventory of systems developed internally, third-party tools, embedded AI solutions and autonomous agents. Each system should have a clearly defined purpose, an accountable owner, a risk classification and a record of the decisions it influences.

This becomes particularly important as employees and business units may begin using AI tools faster than central IT and governance teams can track them.

  • Measure: Make risks measurable

Not every AI application carries the same level of risk. A tool suggesting a marketing headline should not face the same controls as a system influencing a person’s access to credit.

Organisations need to establish measurable thresholds before deployment. These may include acceptable error rates, fairness indicators, security requirements and conditions under which an AI system or agent must be restricted.

Risk classification should influence actual organisational decisions rather than remain limited to documentation.

  • Monitor: Track performance after deployment

Governance cannot end once a system becomes operational. Organisations need to monitor performance changes, error patterns, unusual behaviour, security incidents, user complaints and outcomes across different groups.

Monitoring should also cover the complete AI system, including its data sources, retrieval tools, connected applications, permissions and human workflows.

For autonomous agents, it is particularly important to understand what actions they can perform, which systems they can access and whether their actions can be reversed.

  • Intervene: Ensure the ability to act

Identifying a problem is useful only when an organisation has the authority and processes to respond. Clear escalation mechanisms, predefined thresholds and assigned decision-making responsibilities are therefore essential.

Depending on the nature of the risk, intervention may involve additional human review, restricting system permissions, rolling back to an earlier version or suspending a feature altogether.

Human oversight must also be meaningful. Placing an employee in front of an AI-generated recommendation and expecting automatic approval does not constitute effective supervision. People need the information, competence, time and authority to question or stop a system when required.

  • Learn: Use failures to improve systems

Every incident, near miss, user complaint and unexpected outcome should contribute to improving the system. This requires structured incident reporting, root-cause analysis and feedback from affected users.

Lessons from one deployment should also inform procurement, model selection, product design and organisational controls. In this way, governance becomes a continuous learning process rather than a documentation exercise.

Applying Responsible AI to Public Services

The importance of this approach becomes clearer when applied to public-sector services. Consider an AI assistant designed to help citizens access government schemes and services.

A conventional assessment may focus on privacy safeguards, technical documentation and the accuracy of its initial responses. An insight-driven approach would examine the system throughout its operation.

It would identify which government services the assistant can access and what decisions it can influence. Its accuracy would be measured across different languages and user groups, while failed interactions, incorrect guidance and citizen complaints would be monitored regularly.

If error rates cross predefined thresholds, the system would trigger an appropriate intervention. Feedback would then be used to improve the model, its data and the service itself.

Such an approach is particularly relevant in a country as diverse as India, where language, access and local contexts can influence how people experience digital public services.

India’s Opportunity to Build Practical Governance Models

India’s AI opportunity extends beyond developing advanced models. It also lies in deploying useful, trustworthy and inclusive applications across public services, healthcare, finance, education and other sectors.

The country’s national AI governance framework, built around seven guiding principles and six governance pillars, provides an important starting point. International developments, including the European Union’s AI Act and the risk-based approaches being developed by the US National Institute of Standards and Technology (NIST), also contribute to the wider discussion on trustworthy AI.

However, frameworks become meaningful only when they influence institutional behaviour. Government departments need to be able to explain where AI is being used and who is accountable for its decisions. Hospitals must be able to demonstrate whether AI-supported processes work adequately for the patients they serve. Businesses need to understand the limitations of their systems and establish clear mechanisms for addressing failures.

India has an opportunity to develop governance approaches that are proportionate to risk, sensitive to local languages and practical for both startups and large organisations. This is particularly important for emerging businesses, which need clear guidance without facing unnecessarily burdensome compliance requirements.

Trust cannot be created through regulation alone. It develops when people can see that a system works, understand its limitations and know that someone is accountable when something goes wrong.

Making Accountability Part of the AI Lifecycle

As AI systems become more capable, responsible adoption will depend on more than policies and compliance certificates. Organisations must be able to identify where their systems operate, understand the decisions they influence and detect when their performance begins to deteriorate.

Equally important is having the authority and willingness to intervene when evidence points to a problem. Without this ability, accountability remains incomplete, regardless of how detailed the governance framework may be.

The larger goal is not simply to demonstrate that an organisation is monitoring its AI systems. It is to ensure that it understands what is happening, recognises emerging risks and responds in time. This is what turns governance from a formal control function into an intelligent, continuous process.

Insights shared by Suman Dash, Chief Operating Officer, ACSEL Technology Forum, during the World AI Summit 2026, organized by Elets Technomedia, held on 14-15th October, 2026, in Bengaluru.

SA
Written by

Sahaj Anand

Elets News Network reports on governance, public policy and digital government across India.

IAS Transfer Tracker

Live — 2,069 transfers tracked Updated 18-11-2026, 14:46

The eGov Weekly Briefing

A weekly round-up of governance news, interviews and policy analysis — in your inbox.