Deployment of Proprietary AI Model for Audit Support
The Comptroller and Auditor General (CAG) of India is scheduled to launch an in-house large language model (LLM) on Monday, aimed at assisting auditors in the analysis of complex financial data. Developed exclusively for the watchdog by the Indian Institute of Technology (IIT) Madras, the platform, designated as LLM 2.0, is designed to consolidate years of internal inspection reports, auditing standards, and regulatory guidelines into a single interface.
According to Krishnan Sangaran Subramanian, Deputy Comptroller and Auditor General, the model has been trained on approximately 65,000 internal inspection reports produced by the agency, alongside its established auditing standards. The platform is intended to function as an assistant to auditors, allowing them to input simple queries to retrieve information from the ingested data.
Functional Capabilities and Data Security
The LLM 2.0 is expected to feature three distinct streams tailored to specific auditing functions: compliance audit, financial audit, and works audit. Once financial statements, auditor reports, and board bills are uploaded, the model will perform numerical calculations, conduct analysis, and identify patterns.
Subramanian noted that the technology is intended to automate routine tasks, serving as an aid for preliminary groundwork rather than a replacement for human auditors. “We still have not reached a stage in which it will do an audit. It will be an aid,” Subramanian stated, clarifying that the model is not intended to conduct the audit itself.
To address concerns regarding data integrity and security, the Deputy Comptroller and Auditor General stated that users will only be permitted to ask questions based on the underlying reports powering the model. The system is designed so that auditors cannot upload external data onto the platform or the cloud.
Shift Toward Data-Driven Audit Planning
The introduction of the AI model coincides with a broader shift in the CAG’s methodology. The watchdog is moving away from conventional audit plans based on specific units and is instead adopting a data-driven approach for the 2026-27 cycle. This new method utilizes data from multiple departments to identify notable patterns and red flags.
In state-level audits, for example, the agency is analyzing data related to direct benefit transfers under social sector schemes, public works, and digital salary payments to officials to identify irregularities.
Integrated Audit Progress
The CAG is also nearing the tabling of its initial integrated reports. These audits, which have been in development for one to two years, involve multiple stakeholders to increase efficiency. Current developments include:
- Social Sector Schemes: Integrated performance audits of seven social sector schemes are nearing presentation.
- Urban Governance: Audits concerning “ease of doing business” and “ease of living” are being conducted in 101 cities with populations exceeding 500,000.
- Future Scopes: The agency is currently undertaking integrated audits for multimodal transport and foundational literacy. Plans are also in place for subsequent reports focusing on school dropouts and skill upgradation.
