CAG Must Match AI-Driven Audit Rhetoric with Accountability
CAG Office India (Image official website)
As the CAG projects itself as a digital audit leader in INTOSAI, its credibility hinges on timely, nationwide audits that expose fraud, curb misuse, and deliver value to citizens.
By P SESH KUMAR
NEW DELHI, September 14, 2025 — INTOSAI’s adoption of a cybersecurity and remote audit work plan for 2026–28, with India’s CAG at the forefront, is being touted as a landmark. Large language models trained on audit reports, global AI/ML training programmes for auditors, and GIS-enabled monitoring of infrastructure projects were all showcased as futuristic tools of accountability. The optics are dazzling. But the real question remains: has this translated into tangible value for India’s citizens?
What has the CAG actually demonstrated in terms of exposing fraud, plugging misuse of funds, improving governance, or producing value-for-money audits in everyday schemes that affect millions? At a time when routine all-India performance audits of social welfare, poverty alleviation, health, drinking water, sanitation, employment and ease of doing business remain delayed or absent, the danger is that the constitutional mandate of assessing government performance risks being overshadowed by headline-grabbing experiments in AI and big data.
The Tech Rhetoric
The recent INTOSAI plan makes no small promises. Cybersecurity audits, IoT-enabled remote monitoring, and knowledge building for the 2026–28 cycle were front-paged. India’s CAG led the charge, unveiling its IIT-Madras partnership on a large language model trained on audit and inspection data, announcing a global nine-month AI/ML certification course for auditors, and showcasing PM GatiShakti as proof of big data and GIS tools in monitoring infrastructure. These are the kind of announcements that delight international delegates and make for impressive press headlines. They also raise the expectation that supreme audit institutions can reinvent themselves for the digital age.
What Has Been Achieved
CAG is not starting from zero. Its Compendium on Responsible AI already recorded experiments where machine learning exposed fake beneficiaries in PM-Kisan and housing schemes, or detected repeated use of photographs to claim subsidies. In audits of scholarship schemes, ghost institutions and ineligible recipients were identified. Audit officers have published technical pieces on how clustering algorithms, anomaly detection and network analysis can support scrutiny of large datasets.
The “AI Strategy Framework” laid out a roadmap for embedding such techniques across the audit cycle. There is no doubt that intent has been declared and pilots attempted.
A handful of audit reports have even flagged digital interventions. The 2022 audit of Direct Benefit Transfer schemes mentioned analytics being applied to sift beneficiary records; certain state audits of urban housing have noted duplicate claims exposed by image recognition tools. These are important beginnings. But they remain isolated, rarely scaled across the Union and states, and rarely followed up with data on recoveries, blocked payments or prosecution.
Where the Gaps Lie
The problem is stark. Beyond scattered pilots, there is no evidence yet that AI, IoT or big data have fundamentally changed the way CAG delivers on its constitutional mandate. No major all-India performance audit has yet shown how these tools uncovered systemic fraud worth thousands of crores or prevented large-scale leakage. Meanwhile, bread-and-butter performance audits that should be the backbone of accountability—covering social welfare, poverty alleviation, employment generation, health, drinking water and sanitation, even ease of doing business—are either missing altogether or delayed for years. The last nationwide performance audit of health programmes, for instance, is years behind reality. Employment schemes like MGNREGA or skill-development initiatives have not been subjected to rigorous, timely, technology-driven audits in a manner that convinces the public that oversight is keeping pace with governance.
This creates an uncomfortable imbalance. On one hand, the CAG is making headlines about LLMs, IoT and AI. On the other, it risks diluting its traditional, constitutionally enshrined role of assessing how governments actually perform in delivering basic services to citizens. Stakeholders may fairly ask whether global optics are being prioritised over local accountability.
What Needs to Be Proved
If the CAG is serious about walking the talk, it must demonstrate in black and white where these new tools have made a difference. Which schemes have been audited using AI at scale? How much fraudulent expenditure was detected and recovered? Where has IoT-based monitoring prevented cost overruns in infrastructure? What value has been delivered in rupee terms relative to the cost of deploying these technologies? And alongside these answers, it must release a clear calendar for long-delayed nationwide performance audits in health, welfare, employment and sanitation. Only then will citizens believe that technology is not an escape route from difficult constitutional responsibilities, but a genuine tool to strengthen them.
India’s CAG has positioned itself as a global leader in INTOSAI, projecting an image of a digital, AI-powered audit office. That image will quickly fade if it is not anchored in the reality of robust, timely, nationwide audits of programmes that matter most to citizens. Without clear evidence of value realised, frauds exposed, misuse stopped and lives improved, the rhetoric risks being dismissed as mere optics. The CAG owes it to the people of India—its most important stakeholder as per its own mission and vision statement—to show not only that it can experiment with futuristic tools, but that it can still deliver on its primary constitutional role: holding governments accountable for performance in the sectors that shape daily life.
(This is an opinion piece, and views expressed are those of the author only)
Follow The Raisina Hills on WhatsApp, Instagram, YouTube, Facebook, and LinkedIn
About The Author
