CAG bets on AI, cyber audits and sovereign LLM to enhance public scrutiny

CAG officials said the institution has adopted a formal AI strategy framework making the Supreme Audit Institution (SAI) of India one of the few globally with a published AI roadmap

At the India Artificial Intelligence (AI) Impact Summit being held in Delhi, the Comptroller and Auditor General (CAG) of India today laid out a roadmap that moves audit from sample checks and manual review to data-driven, AI-enabled assurance at scale.

 

Speaking at a session on “Leveraging Artificial Intelligence in Public Audit for Greater Transparency and Accountability”, the institution unveiled its plans to experiment with drones, computer vision and high-resolution satellite imagery to verify on-ground realities in areas such as mining leases and scheme implementation.

 

Underpinning these efforts is a plan for a “data lake house” to pool structured and unstructured data, and a sovereign large language model (LLM) for audit being developed with IITs, alongside a major skilling drive that aims to train about 5,000 officers in data science, AI and cyber security.

 

 

CAG officials said the institution has adopted a formal AI strategy framework, released in April last year, making the Supreme Audit Institution (SAI) of India one of the few globally with a published AI roadmap.

 

The framework rests on four pillars: using AI and machine learning in core audit and internal business processes, auditing AI systems deployed by government departments, large-scale capacity building of auditors and sustained research and development backed by enabling infrastructure.

 

Building on two decades of information systems (IS) audits on flagship platforms such as the income tax system, GST and IRCTC, the CAG is now pushing beyond traditional IT audits into cyber security assurance.

 

“Given the scenario of a lot of cyber attacks, we felt that over and above the certain empanelled audits we are going to do an annual audit on the identification and give an assurance on the cyber security of an application,” said K Surjith, director, office of the CAG.

 

The session showcased several concrete AI use cases already reshaping audit practice. For beneficiary-heavy welfare schemes, where tens of crores of people are enrolled, the CAG is developing an “audit toolkit” that allows even non-technical staff to feed in beneficiary images and flag duplicate or ghost entries, backed by big data analytics on entire beneficiary databases rather than limited samples.