AI Audit Cracks Three-Year Police Salary Scam in Chhattisgarh

An AI audit of police salary records in Chhattisgarh’s Bastar district has led to the arrest of three constables over an alleged fraud of between Rs 1.5 crore and Rs 2 crore, police said on Wednesday. The audit, which examined data on around 2,000 police personnel, is the first time artificial intelligence tools have been used in such an exercise, according to investigators. The trio are accused of siphoning government funds in small monthly increments over nearly three years.

The accused, constables Girish Rai, Rajkumar Katlam and Hemant Mathew, allegedly manipulated digital salary records in the office of the Superintendent of Police in Jagdalpur between October 2023 and May 2026, Bastar Superintendent of Police Shalabh Kumar Sinha said. Rai worked as an assistant in the salary section and is believed to be the ringleader, Sinha said. A case has been registered under sections of the Bharatiya Nyaya Sanhita covering cheating, forgery, criminal breach of trust and embezzlement of government funds. All three have been remanded to 14 days of judicial custody.

Three Arrests in Chhattisgarh’s Bastar

The arrests follow an internal audit of police salary data in Bastar, where AI tools were used to comb through payroll records for unusual patterns, officials said. The audit was conducted by a DSP-rank officer assigned to the case, after the AI flagged anomalies that manual checks had missed. A case was registered at Jagdalpur police station on Monday before the trio was taken into custody.

The scale of the alleged theft is the largest single police salary fraud publicly disclosed in Chhattisgarh in recent memory, according to the case filings reviewed by reporters. Investigators put the loss at between Rs 1.5 crore and Rs 2 crore, siphoned off in small monthly increments across nearly three years. Rai had been posted in the SP office since 2012, on compassionate grounds, and his long tenure in the salary section gave him direct control over how the bills were prepared, Sinha said. The use of AI in the audit is part of a wider shift in how government payrolls are being scrutinised.

Katlam and Mathew were posted in different sections of the same office and are believed to have benefited from the inflated payouts without handling the records themselves. The probe is still in its early stages, and police have not ruled out the possibility of further arrests.

The Method Behind the Alleged Salary Manipulation

Investigators say the case centres on what they describe as a sustained, methodical manipulation of police payroll records in the Jagdalpur office of the Superintendent of Police. The inquiry has identified two distinct mechanisms, the first targeting the three accused’s own salaries, the second involving other employees. The probe is still mapping the full extent of the alleged scheme. Initial findings suggest the manipulation was carried out in small monthly increments designed to look like normal payroll adjustments.

Police are now treating the wider pool of employees as potential witnesses rather than suspects at this stage. Officials said the records show that Rai had been in the salary section long enough to know which approvals attracted the least scrutiny. The investigation will also examine whether the same method was used to siphon funds from budgets other than salaries. A DSP-rank officer is leading the case with support from the district audit team. Investigators say the trial proceedings could take several months given the volume of financial records under examination.

The alleged method, as described by investigators:

• Edited soft copies of digital salary records before they were finalised
• Inflated payouts for the three constables in small monthly increments over nearly three years
• Diverted the excess into personal or controlled bank accounts
• Inflated loan amounts for other employees and collected the difference in cash as repayment

The AI Audit That Caught the Fraud

The case was uncovered during a departmental audit that turned to AI tools when the volume of salary data made manual review impractical. The audit processed records covering around 2,000 police personnel in the Bastar range, officials said. The system flagged an unusual rise in salary expenditure that pointed investigators to the three accused.

Bastar Superintendent of Police Shalabh Kumar Sinha said the data volume drove the decision to deploy AI tools. The use of AI was a first for such an audit, according to Indian Express, and the tools isolated patterns that conventional sample-based reviews had missed. Once the AI surfaced the abnormal payment patterns, the DSP-rank officer assigned to the case worked back through the records and the bank statements. The methodology is now being examined for wider deployment across other government salary systems.

The data related to salaries was voluminous and so auditors proactively decided to use AI tools.

Officials associated with the investigation said identifying such subtle irregularities through conventional manual scrutiny would have been extremely difficult given the volume of data. They said the AI-based analysis directed investigators towards suspicious transactions, rather than catching the fraud on its own. The shift mirrors a wider trend, with the US Treasury already using AI to detect check fraud in near real time and reclaim stolen funds.

The pattern AI is now being used to flag is exactly the kind that legacy audit methods miss: small, repeated, distributed across multiple accounts. Government payrolls are a particular blind spot because, unlike project-based spending, they are not subject to line-by-line financial audits. The Bastar case has put that gap in focus for other states too. Sinha said the AI tools will be considered for use in future audits of government salary systems in the district.

Why It Stayed Hidden for Nearly Three Years

Investigators say the alleged manipulation survived for nearly three years in part because of how police payrolls are normally structured. Salary budgets, unlike project-based allocations, are not subject to line-by-line financial audits in the same way. The accused also spread the alleged theft across three salaries rather than concentrating it in one, which avoided triggering a single-account spike alert. Small monthly increments stayed well below the threshold that would have prompted a manual review. The pattern is, investigators say, the kind that legacy audit methods consistently miss.

The fraud went undetected because police salary expenditure fluctuates frequently due to regular transfers, postings and changes in personnel strength. The accused allegedly inflated salaries by small amounts each month in their own names and those of the other two constables, allowing the fraud to remain unnoticed for years.

The investigation is also going back through older salary records to check whether the same method was used at other police stations in Bastar. Officials said the audit team is being expanded to handle the additional workload.

Investigators are now going through the digital records again to reconstruct the full pattern of alleged theft. They are also checking whether the same method was used to siphon funds from budgets other than salaries. The case is being treated as the first of what could be a wider review.

AI Audits Are Spreading Across Government Payrolls

The Bastar case sits inside a much wider shift in how government payrolls are being audited. AI and machine learning are increasingly being deployed to identify fraudulent transactions, improper payments and program abuse across federal, state and local systems, according to a Brookings analysis of AI in government fraud detection. The pattern is the same one fraud-detection systems in the private sector have used for years.

AI fraud-detection systems work by scanning huge transaction data sets for patterns that human reviewers would miss. In the private sector, banks and payment processors have used these systems for over a decade to flag unusual account activity. Government adoption is now catching up, though unevenly, and salary systems are one of the more obvious places to start.

The shift is also visible in the audit profession itself: two-thirds of audit firms now run AI tools, though most still require human validation before any finding is finalised. Salary fraud cases are the kind of problem where that combination works well, because the AI surfaces the anomalies and the auditor confirms them.

The same algorithms used to flag suspicious transactions can be repurposed for benefits, payrolls and grant programmes. The US Department of the Treasury, for one, has built an AI system to detect check fraud in near real time. State governments are now exploring whether the same playbook can be used to catch the kind of small, distributed salary inflation the Bastar case turned on.

What Investigators Are Tracking Next

Documents related to salary payments, bank accounts and financial transactions are being examined to identify other possible beneficiaries or officials linked to the alleged scam. The probe is being led by a DSP-rank officer with support from the district audit team.

Officials said the investigation will also look at the digital audit trail to see whether the AI tools can be run against other government salary systems in Chhattisgarh. A broader push for AI-assisted audits of government payrolls is under discussion, they added, after the case demonstrated how conventional checks can miss small, repeated irregularities. The case has renewed discussions on strengthening financial oversight and digital auditing within government departments. For now, the three accused remain in judicial custody, and the auditors are combing through records that may show whether the alleged scheme reached beyond the Bastar SP office. The trial is expected to take several months given the volume of financial records under examination.