Maharashtra’s power distribution utility says it detected 27,964 electricity theft cases and issued assessment bills worth Rs 66.28 crore during a three-month campaign, using artificial intelligence to identify suspicious consumption patterns before sending enforcement teams to inspect locations. The figures point to a broader shift in how a public utility is attempting to manage commercial losses: from predominantly survey-based inspection towards a combination of automated risk identification, field verification and digital case tracking.
Maharashtra State Electricity Distribution Company Limited, or MSEDCL, announced the action in an official statement on September 10. The campaign covered June to August 2026 and involved more than 190 flying squads drawn from the utility’s Operations and Maintenance, Security and Enforcement divisions. The company said the squads acted on leads generated by an AI and machine-learning module designed to identify possible electricity theft.
The intervention is significant not simply because of the number of cases recorded, but because it links three distinct parts of electricity governance. The first is detection: identifying accounts or locations whose consumption patterns appear unusual. The second is physical verification: sending personnel to inspect meters and connections. The third is enforcement administration: recording inspection findings, assessing penalties, disconnecting supply where applicable and pursuing recovery. MSEDCL’s account suggests that these functions are increasingly being integrated into one digital workflow.
The utility said its AI module analyses more than 50 technical parameters around the clock. These include historical electricity consumption, sudden increases or decreases in demand and other anomalies that may indicate meter tampering. Consumers displaying unusual patterns are classified as high-risk, allowing enforcement teams to focus on selected locations rather than beginning with extensive manual surveys.
That process does not, according to the information supplied by MSEDCL, establish theft by itself. The algorithm generates leads, while physical inspections determine whether enforcement action is warranted. This distinction is central to the system’s credibility. A change in consumption can have several explanations, including alterations in occupancy, seasonal demand, business activity or a technical problem. The reported workflow therefore places field inspection between automated suspicion and formal assessment.
MSEDCL said the campaign resulted in assessment bills totalling Rs 66.28 crore in 24,508 verified cases. The difference between the 27,964 cases detected and the 24,508 cases for which assessment bills were issued indicates that not every initial detection had reached the same stage of verification or billing by the time of the announcement. The source does not provide a breakdown of the remaining cases, so the reasons for that gap cannot be established from the available information.
The utility reported that it had recovered Rs 29.85 crore from 10,810 settled cases. This represents the amount recovered from cases described as settled, rather than the total value of all assessments issued. The figures also show that detection and recovery are separate stages of the enforcement process. Identifying a suspected case may be accomplished through data analysis, but converting it into recovered revenue requires inspection, assessment, payment or compounding, and in some cases legal proceedings.
The regional distribution of cases was uneven. The Konkan region recorded the highest number, with 11,083 cases. Nagpur reported 5,949 cases, Pune 5,849, and Chhatrapati Sambhajinagar 5,083. Together, these figures account for all 27,964 cases reported during the campaign. The source does not provide the number of electricity connections, consumer categories, geographical coverage or inspection intensity in each region. As a result, the regional totals cannot by themselves be used to compare the underlying prevalence of theft between regions.
MSEDCL has also digitised the enforcement process through its Theft Tracking and Monitoring System mobile application. According to the utility, inspection details, live photographs, digital panchnamas, immediate disconnection records and penalty recovery information are uploaded in real time. The stated purpose is to improve transparency and reduce the scope for negligence during enforcement action.
Digitisation changes the administrative record of an enforcement visit. Instead of relying only on paper files and later data entry, the utility says information can be captured at the location and monitored through the case lifecycle. In principle, this creates a traceable sequence from the initial lead to inspection, assessment and recovery. However, the supplied material does not establish how the system is audited, who can access the records, how disputes are handled or whether consumers can review the evidence used against them.
Those questions matter because electricity enforcement operates at the intersection of a basic utility service and statutory penalties. MSEDCL said consumers who do not pay assessed charges and applicable compounding fees within the stipulated period may face legal action. It reported that 289 criminal cases had been registered under Section 135 of the Electricity Act against alleged offenders who had not cleared their dues. The company also said electricity theft can attract financial penalties and imprisonment of up to three years, depending on the circumstances.
The use of the word “alleged” is important in describing these cases. An AI-generated risk classification is not the same as a legal finding, and an assessment bill is not necessarily equivalent to a completed prosecution. The information supplied by the utility does not include conviction data, the number of cases challenged by consumers, or the outcome of the 289 criminal cases. It therefore supports an account of enforcement activity, but not a conclusion about the final legal status of all detected cases.
MSEDCL’s campaign is also tied to a broader financial target. Lokesh Chandra, Additional Chief Secretary to the Chief Minister and MSEDCL’s chairman and managing director, said the company is aiming to reduce Aggregate Technical and Commercial losses, or AT&C losses, to below 10 per cent over the next three years. AT&C losses generally capture both technical losses in the distribution network and commercial losses associated with billing, collection and unauthorised consumption. In the supplied statement, MSEDCL presents electricity theft enforcement as one part of its effort to improve this measure.
The source does not provide MSEDCL’s current AT&C loss figure, a baseline for the target, or a separate estimate of how much of the loss is attributed to theft. That absence limits what can be concluded about the likely contribution of the AI programme to the three-year objective. The campaign’s reported assessment and recovery figures demonstrate enforcement activity, but they do not yet show how the intervention has changed overall distribution performance.
The institutional model described by MSEDCL combines centralised data analysis with decentralised field action. A technical module monitors consumption information, while flying squads operate across the utility’s four major regions. The arrangement reflects a practical division of responsibilities: data tools narrow the search, and field teams establish the facts needed for administrative or legal action. The mobile application then creates a common record for the enforcement process.
This model may allow a utility to deploy inspection capacity more selectively. More than 190 squads cannot physically inspect every consumer connection at the same frequency, particularly across a large state distribution network. A system that prioritises locations based on multiple parameters is intended to direct scarce enforcement resources towards cases with stronger indicators of irregularity. Whether it improves accuracy depends on the quality of the underlying consumption data, the design of the risk model and the reliability of the subsequent inspection process. The source confirms the intended use of the system, but does not provide an accuracy rate or independent evaluation.
The campaign also raises a governance question about the use of automated suspicion in public services. MSEDCL says the system examines more than 50 parameters, but the statement does not identify the relative weight of those parameters, the threshold for classifying a consumer as high-risk or the safeguards against repeated targeting. Nor does it specify whether consumers receive an explanation of the evidence supporting an assessment. These details are not necessary to establish that the campaign took place, but they are necessary to assess whether data-led enforcement is fair, accountable and proportionate.
For citizens, the immediate effect of the programme is the possibility of more targeted inspections, faster disconnections in cases where the utility says violations are established, and greater use of digital records in assessment and recovery. For MSEDCL, the reported recovery of Rs 29.85 crore from 10,810 settled cases provides an early measure of financial collection, while the remaining assessed amount shows that detection does not automatically translate into settlement. The supplied material does not explain the payment timelines, appeal mechanisms or consumer assistance available during disputes.
The evidence currently confirms four developments: MSEDCL used AI and machine learning to generate electricity-theft leads; more than 190 flying squads followed up those leads; the utility recorded 27,964 cases during June-August 2026; and its digital application was used to document enforcement actions. It also confirms substantial assessment, recovery and criminal-case figures reported by the utility. What remains uncertain is the system’s independently measured accuracy, the outcome of disputed cases, the contribution of theft reduction to total AT&C losses and the effect of the campaign on service quality or consumer behaviour.
The next stage of scrutiny will therefore be whether MSEDCL publishes fuller performance data as it pursues its below-10-per-cent AT&C loss target. The most useful indicators would include verified cases by consumer category and connection base, recovery rates, legal outcomes, appeals, the treatment of false positives and changes in overall distribution losses. Until those details are available, the campaign is best understood as a significant administrative shift towards data-assisted electricity enforcement, rather than conclusive evidence that artificial intelligence has solved the utility’s commercial-loss problem.
