The Economic Times reported Oct. 29 that the Comptroller and Auditor General of India (CAG) is seeking wider use of artificial intelligence (AI) in audits, including analysis of whole transaction sets for risks and control weaknesses.
The Economic Times said the recommendation appears in the CAG’s 2025–26 public-governance report and concerns AI, machine learning and transaction analytics. It reported that the document urges reviews of complete transaction sets and use of data from digital platforms. The newspaper quoted the report as saying digital information would support “more comprehensive and evidence-based assurance.”
The CAG’s technology program predates the report described by The Economic Times. Its April 2025 AI strategy framework proposes using AI to review data, detect unusual patterns, assess risks and produce predictive insights during audit work. Comptroller and Auditor General K. Sanjay Murthy signed the document, which also identifies data privacy, model governance, responsible use and staff capacity as adoption considerations.
The CAG audits the accounts of India’s Union and state governments, as well as public-sector organizations. The CAG says its authority derives from Articles 149 and 151 of India’s Constitution and the Comptroller and Auditor General’s (Duties, Powers and Conditions of Service) Act, 1971.
In a Sept. 17, 2025, release, the CAG described risk models drawing on data repositories and past audit findings, a large language model in development, and data analytics cells in field offices. The CAG said Tamil Nadu gave it access to over 20 databases and key government IT systems.
A government eProcurement System listing records a May 2026 CAG tender titled “Sovereign AI and Data Platform with Agentic AI Applications.” It gives June 30, 2026, as the bid-opening date and shows no award result.
CAG audit regulations say analytics can identify exceptions, trends and deviations, while auditors must consider data reliability and controls. The rules also allow for field examination or sample verification when needed.
