内容来源：https://news.microsoft.com/source/asia/features/a-race-against-time-maharashtra-police-get-an-ai-copilot-to-fight-cybercrime/

内容总结：

印度马哈拉施特拉州引入AI助手打击网络犯罪，调查效率显著提升

面对日益猖獗的网络诈骗，印度马哈拉施特拉州警方引入了一款名为“MahaCrimeOS AI”的人工智能辅助调查平台，显著提升了案件处理效率与规范性。该平台由州政府与微软合作开发，目前已从试点地区推广至全州1100个警察局。

AI助手应对破案压力
近年来，印度网络犯罪案件激增。根据印度网络犯罪协调中心（I4C）的数据，仅2024年就报告了近227万起案件，经济损失超过2280亿卢比。面对海量案件与复杂的技术追踪工作，基层警力常感力不从心。

银行职员尼图遭遇的虚假投资应用诈骗案是其中一例。她向诈骗应用转账超过300万卢比后察觉异常，随即报警。在纳格浦尔农村地区的萨奥内警察局，借助MahaCrimeOS AI系统，警方仅用15分钟就完成了立案登记。过去，仅收集初步调查信息就可能耗时两三个月。

“数字搭档”赋能一线警员
该平台的核心是一个“调查副驾驶”AI助手。它能自动解析报案人上传的各类证据（如截图、银行流水），提取关键信息并归类；根据案件类型，自动生成标准化法律文书（如向银行调取账户信息的函件）；还能接入开源情报数据库，辅助分析案件关联性与嫌疑人定位。

“以前处理一个案件要手工起草多份文书，往返多个银行，极易出错遗漏，”助理警察督察阿西什·辛格·塔库尔表示，“现在点击几下就能批量调取通话记录，系统自动生成函件，我每月能处理七八个案件，而以前只能处理一个。”

标准化与智能化并重
平台开发方CyberEye公司指出，初期挑战在于警务流程缺乏标准化——同类诈骗的描述用语不一，调证文书格式各异。通过与州政府专设的“MARVEL”项目合作，团队将警方的调查规程与当地语言（马拉地语）集成到平台中，确保了操作的规范性与易用性。

纳格浦尔农村地区警察局长兼MARVEL首席执行官哈什·A·波达尔强调，该平台不仅能固化现有最佳调查实践，还能针对具体案件生成个性化的调查计划，相当于为每位警官配备了一位融合了警方集体经验的“数字导师”。

提升警员信心与案件关注度
对于长期处理网络犯罪的助理调查员桑吉塔·加万德而言，平台极大地减轻了行政负担。“以前被问及案件进展时，我可能因文书工作繁重而无法及时回应。现在我能更快完成报告，有时问审视每个案子，压力小了，信心也更足了。”她表示。其所在警局处理的月均案件量已从数年前的两三起增至如今的约两百起。

马哈拉施特拉邦首席部长德文德拉·法德纳维斯指出，此项合作始于应对网络犯罪，但其“合乎伦理、负责任地为公共利益服务”的AI应用模式，有望拓展至医疗、农业、治理等多个领域。

对于受害者尼图而言，尽管被骗资金尚未全部追回，但警方的有力支持给了她慰藉。“经济损失带来的痛苦日渐消退，但警察的支持永远不会褪色，”她说。目前，警方已通过冻结全国多个账户，为其追回了约80万卢比。

中文翻译：

与时间赛跑：马哈拉施特拉邦警方引入AI助手打击网络犯罪

银行职员妮图·Y（Nitu Y）向一款在线股票交易应用投入资金约一个月后，开始怀疑事情有些不对劲。

根据该应用显示，她通过所谓购买的股票获得了900万卢比（约合10.2万美元）的利润。当她试图提取这些利润时，对方却告知她需要先支付数十万卢比的佣金。

她转账了这笔钱。但与她交涉的人坚称未收到款项。这时，她的银行业知识和直觉起了作用。

"他们告诉我，RTGS（银行间实时全额结算系统）不是实时到账的……那一刻我突然意识到，这可能是欺诈。事情不对劲，"妮图说。

得益于她的工作，她知道要立即拨打1930，这是印度网络犯罪协调中心（I4C）下设的全国网络犯罪举报热线。

她也很幸运地居住在印度西部马哈拉施特拉邦的那格浦尔地区。

自今年四月起，当地警方一直在使用"MahaCrimeOS AI"平台。这是一个基于微软Foundry平台定制的犯罪调查系统，能帮助警方更快地处理投诉，并梳理复杂的数据和流程——这些都是处理网络犯罪的关键功能。

2025年12月12日，马哈拉施特拉邦政府和微软宣布，MahaCrimeOS AI将从那格浦尔的23个警察局推广至全邦1100个警察局。

"为公共利益开发合乎道德、负责任的AI是我们的核心理念，"马哈拉施特拉邦首席部长德文德拉·法德纳维斯表示，"人工智能有能力通过提高效率、改善生活质量、为每位公民带来真正的便利生活来实现变革。"

该邦甚至为此成立了一个特殊目的机构，以在执法领域引入AI解决方案。该机构缩写为MARVEL，全称为"马哈拉施特拉邦先进研究与改革法律执行监督机构"。"创建MARVEL背后的愿景是建立一个平台，与全球深度技术领导者合作，开发能够从根本上改变我们治理方式的AI助手，"法德纳维斯说。

近年来，随着投资骗局、退款欺诈和"数字逮捕"等案件的激增，妮图的案件只是印度数百万网络犯罪投诉中的一例。根据2025年12月2日一份议会答复中引用的I4C数据，仅2024年就报告了近227万起网络犯罪案件。当年，网络犯罪分子造成的损失超过2280亿卢比（约25亿美元）。I4C是由印度内政部设立的国家机构，旨在为印度预防、侦测、调查和起诉网络犯罪建立一个框架。

由于大多数调查员并不精通技术，手头有一个AI驱动的数字助手减轻了他们的工作负担。

在妮图的案件中，在那格浦尔的萨奥内尔警察局提交刑事投诉（即"第一信息报告"，FIR）只花了15分钟。

她提供了诈骗网站的社交媒体账号和网址、手机号码、短信截图和银行对账单。所有这些信息，无论是PDF文件还是手写的英文、印地语或当地马拉地语笔记，都能被MahaCrimeOS AI在几分钟内上传、提取并归档到相应部分，使调查员从繁琐的手工创建案卷工作中解放出来。

处理妮图案件的助理警察督察阿希什·辛格·塔库尔表示，以前需要两到三个月来收集调查所需的信息。

"我们得跑多家银行，因为资金被转移到了超过40到50家不同的银行……我们必须给他们发函并获取他们的说明。此外，还涉及多个手机号码……我得获取IP地址。很多都是手工活，"他解释道。

而现在，塔库尔只用了一周左右就完成了这些任务并收到了这些服务提供商的回复。

"只需点击几下，我就可以申请调取多个号码的通话详单。我不必起草信函；它们已经由MahaCrimeOS AI草拟好了。我也不用担心遗漏信息或在信函中出错，然后不得不全部重来，"他说。

他现在每月能处理七到八起案件，而以前只能处理一起。

助力执法的工具

实现这些一线流程的自动化只是挑战的一部分，为MahaCrimeOS AI提供引擎支持的独立软件供应商CyberEye的创始人兼首席执行官拉姆·加内什表示。该公司与MARVEL合作，为马哈拉施特拉邦警方定制了该平台。CyberEye是微软合作伙伴ISV，并与微软印度开发中心紧密合作。

"我们面临的第一个挑战是标准化程度极低，"他说。

这包括在报告中使用不同的术语来描述相同的金融欺诈或犯罪手法，以及没有向银行索取信息的标准法律通知格式。

这使得警方更难决定下一步行动、发现案件是否关联或积累传承性知识。

CyberEye与MARVEL合作，整合了马哈拉施特拉邦警方的调查规程，并使平台支持马拉地语，方便调查员使用。

随后，他们以调查助手的形式加入了AI功能，该助手能分析案件信息、生成自动化工作流程并建议调查路径。该助手是使用微软Foundry中的Azure OpenAI服务构建，并通过Microsoft Defender for Cloud确保安全。

该平台还内置了印度刑法和开源情报的访问功能，可用于关联犯罪和定位嫌疑人。

该平台的一个关键优势是能够适应现有的最佳实践规程，添加针对特定犯罪的知识层级，然后制定量身定制的调查计划，那格浦尔农村地区警察局长兼MARVEL首席执行官哈什·A·波达尔表示。

"现在你有了一个助手，它能够利用马哈拉施特拉邦警方内部现有的知识资源来完成这项工作，"他说。

该系统还通过复制和自动化警官日常的行政任务，使他们能更专注于破案和援助受害者。

当调查员登录MahaCrimeOS AI时，他们可以查看所有正在处理的案件，以及是否有新信息或待采取的行动。

在金融欺诈案件中，此类行动包括要求银行冻结账户、请求封停电话号码或下架社交媒体账号。

同时，如果从电信公司收到了通话详单，他们可以询问调查助手如何进行下一步。或者，他们可以点击助手的建议来分析记录或收集关于某个电话号码的开源情报。

这些功能使警官能够为每一起网络犯罪制定非常个性化的调查计划，波达尔说。

MahaCrimeOS AI如何运作

"主要优势之一是调查速度，"他说。"其次是确保调查符合马哈拉施特拉邦警方设定的最高标准。第三是使调查官有能力向受害者解释他们正在采取的步骤。"

该平台还增强了处理网络犯罪警官的自信心和自主性。

据波达尔称，以前一些调查员在面对网络犯罪案件时"几乎僵住了"，等待更资深的同事给予详细指示。现在他们更有信心了。

"现在他们有了一个数字伙伴……这无疑构建了大量的专业能力，进而成为各级警官获得更多自信和做出更好决策的助产士，"他说。

助理调查员桑吉塔·加万德过去七年来一直在那格浦尔农村地区警方从事网络犯罪工作。刚开始时，她每月处理两到三起案件。现在，过去两年里，她每月处理约200起案件。

她过去的主要挑战是应付繁重的行政工作。现在，她更加自信了。

"以前，我可能没时间处理某些案件，"她说。"如果老板问起，我也无法给出答案。现在我更有信心，压力也小了，因为我能更快完成报告，并且有时间查看每个案件。"

打击网络犯罪的持续努力

像MahaCrimeOS AI这样的数字助手，增强了印度打击网络犯罪的力量。联邦政府已实施了诸如1930热线、国家网络犯罪举报门户网站，以及在邦警察部门内设立名为"网络小组"的专门单位等措施，所有这些都由I4C协调。

该平台也为AI更广泛的应用树立了榜样。

"我们与微软的合作始于解决复杂的网络犯罪挑战，但其潜力远不止于此，"首席部长法德纳维斯说。"如今，AI触及人类活动的各个领域，从医疗保健、农业到工业和治理，我们打算负责任地利用这种力量，创建一个更高效、以公民为中心的邦。"

对于银行职员妮图来说，她的磨难仍在继续，但她仍然感谢那格浦尔警方。

在三月到四月期间，她总共向那个虚假应用转账超过300万卢比（约合3.4万美元）。截至目前，警方已从全国范围内冻结的银行账户中追回了约80万卢比（约合9000美元）。

"（警方的）支持比我拿回钱更重要，因为已经过去六个月了，"她说。"我经济损失的严重性正日渐消退，但来自警方的支持永远不会消退。"

顶部图片说明： 对于助理调查员桑吉塔·加万德（中）和她在那格浦尔农村地区警方的同事们来说，拥有MahaCrimeOS AI作为数字伙伴增强了他们处理网络犯罪的信心。摄影：Selvaprakash Lakshmanan for Microsoft。

图片库中描绘的数据完全是虚构的，仅用于说明目的。

英文来源：

A race against time: Maharashtra police get an AI copilot to fight cybercrime
Bank clerk Nitu Y had been putting money into an online stock-trading app for about a month before she suspected something was amiss.
According to the app, she made profits of 9 million rupees (approximately U.S. $102,000) on shares she supposedly had bought. When she tried to cash out those profits, she was told to first pay hundreds of thousands more in commission fees.
She transferred the sum. But the person she dealt with insisted that the money had not been received. That’s when her banking knowledge, and instincts, kicked in.
“They told me that RTGS (the real-time gross settlement system between banks) is not a real time process … Then it clicked in my mind that something is fraudulent. Something is wrong,” Nitu said.
Thanks to her job, she knew to immediately call 1930, India’s national helpline for reporting cybercrime under the Indian Cyber Crime Coordination Centre (I4C).
She was also fortunate to live in Nagpur district in the western Indian state of Maharashtra.
Since April, police there have been using MahaCrimeOS AI, a customized crime investigation platform powered by Microsoft Foundry that helps them process complaints faster and navigate complex data and procedures — all crucial functions for handling cybercrime.
On Dec. 12, 2025, the government of Maharashtra and Microsoft announced that MahaCrimeOS AI will be extended from Nagpur’s 23 police stations to all 1,100 police stations across the state.
“Ethical and responsible AI for public good is our core motto,” said Maharashtra Chief Minister Devendra Fadnavis. “AI has the power to transform by improving efficiency, enhancing quality of life and delivering true ease of living for every citizen.”
The state even created a special purpose vehicle to introduce AI solutions in law enforcement. It is known by the acronym MARVEL, which stands for Maharashtra Advanced Research and Vigilance for Enforcement of Reformed Laws.” The vision behind creating MARVEL is to build a platform that partners with global deep-technology leaders to develop AI copilots that can fundamentally transform the way we govern,” said Fadnavis.
Amid a surge in recent years of investing scams, refund frauds and digital arrests, Nitu’s case is just one of the millions of cybercrime complaints lodged in India. According to I4C data cited in a parliamentary reply on Dec. 2, 2025, nearly 2.27 million cybercrime cases were reported in 2024 alone. More than 228 billion rupees (US $2.5 billion) were lost to cyber criminals that year. I4C is the national agency set up by the Ministry of Home Affairs to create a framework for preventing, detecting, investigating and prosecuting cybercrime in India.
With most investigators not well-versed in tech, having an AI-powered digital assistant on hand has eased their workload.
In Nitu’s case, lodging her criminal complaint – known as a first information report or FIR – took just 15 minutes at the Saoner police station in Nagpur.
She supplied the scam site’s social media account and URL, mobile numbers, screenshots of her text messages and bank statements. All this information, be it on a PDF file or a handwritten note in English, Hindi or local language Marathi, can be uploaded, extracted and filed in their respective sections by MahaCrimeOS AI in minutes, freeing an investigator from tedious manual work to create a case file.
Assistant Police Inspector Ashish Singh Thakur, who handled Nitu’s case, said that before it would have taken two to three months to gather information for the investigation.
“We have to go to multiple banks because the funds were transferred across more than 40 to 50 different banks … We have to write them a letter and take a statement from them. Then, there were multiple mobile numbers used … I have to get the IP addresses. Lots of manual work,” he explained.
Instead, it took Thakur about a week to complete these tasks and hear back from these service providers.
“With a few clicks I can request the call detail records of multiple numbers. I don’t have to draft the letters; they’ve been drafted by MahaCrimeOS AI. And I don’t have to worry about missing out information or making mistakes in my letters, and having to start all over again,” he said.
He can now handle seven to eight cases a month, compared with just one previously.
Tools to aid law enforcement
Automating these frontline processes was just one part of the challenge, said Ram Ganesh, founder and CEO of CyberEye, the independent software vendor (ISV) that built the engine powering MahaCrimeOS AI. The company worked with MARVEL to customize it for Maharashtra police.
CyberEye is a Microsoft Partner ISV and works closely with the Microsoft India Development Center.
“Our first challenge was there was very minimal standardization,” he said.
This ranged from using varying terms in reports to describe the same financial fraud or criminal method, to not having standard legal notices seeking information from banks.
This made it harder for the police to decide on next steps, discover if cases were linked or build up legacy knowledge.
CyberEye worked with MARVEL to incorporate the Maharashtra police’s investigation protocols and make the platform available in Marathi, for ease of use by investigators.
They then added AI, in the form of an investigation copilot that analyses the case information, generates an automated workflow and suggests investigation pathways. The copilot was built using Azure OpenAI Service in Microsoft Foundry and secured with Defender for Cloud.
The platform also has built-in access to India’s criminal laws and open-source intelligence, which can be used to link crimes and locate suspects.
A key advantage of the platform is its ability to adapt existing best-practice protocols, add a level of knowledge specific to a crime and then develop tailored investigation plans, said Harssh A Poddar, superintendent of police, Nagpur Rural, and CEO of MARVEL.
“Now you have a copilot that is able to do that work taking on board knowledge resources that exist within the Maharashtra police,” he said.
The system also frees officers to focus more on crime-solving and victim assistance, by replicating and automating administrative tasks they do in a typical day.
When an investigator logs on to MahaCrimeOS AI, they can view all the cases they are working on, whether there is new information on it or action to be taken.
In a financial fraud case, such actions include asking a bank to freeze an account, requesting a phone number be blocked or that a social media account be taken down.
Meanwhile, if call detail records have arrived from a telecommunications company, they can ask Investigation copilot how to proceed. Or they can click on its suggestion to analyze the records or gather open-source intelligence on a phone number.
These features enable officers to develop a very individualized investigation plan for every single cybercrime, said Poddar.
How MahaCrimeOS AI works
“One of the main advantages has been the speed of investigation,” he said. “The second is the assurance that the investigation is meeting the highest standards set by the Maharashtra police. And the third is enabling the capacity of the investigation officer to explain to the survivor the steps that they are taking.”
The platform has also fostered self-confidence and autonomy among officers handling cybercrimes.
According to Poddar, some investigators “almost froze” when faced with cybercrime cases before and waited for more senior colleagues to give them detailed instructions. Now they are more confident.
“Now they have a digital companion … It certainly builds a lot of professional capability, which in turn serves as a midwife for more self-confidence and much better decision making for officers at all levels,” he said.
Assistant Investigator Sangita Gawande has been working in cybercrimes for the past seven years with Nagpur Rural police. When she first started, she saw two or three cases a month. Now, in the past two years, she has seen about 200 cases monthly.
Her key challenge used to be staying on top of her administrative workload. Now, she’s more self-assured.
“Before, I wouldn’t have the time to get to certain cases,” she said. “And I wouldn’t be able to provide answers if my boss asked about them. Now I’m confident and less stressed because I can complete my reports quicker and have time to look at each case.”.
A continuous effort to fight cybercrime
A digital assistant like MahaCrimeOS AI adds to India’s efforts to fight cybercrime. The federal government has implemented measures like the 1930 hotline, National Cybercrime Reporting Portal, and special units within state police departments called Cyber Cells, all coordinated by I4C.
The platform has also set an example for how AI can be used more widely.
“Our collaboration with Microsoft began with solving complex cybercrime challenges, but its potential is far greater,” said Chief Minister Fadnavis. “AI today touches every sphere of human activity, from health care and agriculture to industry and governance and we intend to harness this power responsibly to create a more effective, citizen-centric state.”
For bank clerk Nitu, her ordeal continues, but she remains grateful to Nagpur police.
In total, she had transferred more than 3 million rupees (about U.S. $34,000) to the fake app between March and April. So far, police have recovered about 800,000 rupees (approximately U.S. $9,000) of that from frozen bank accounts across the country.
“The support is more important than me getting my money back, because it has been six months,” she said. “The intensity of my financial loss has been fading day by day, but the support from the police will never fade.”
Top image: For assistant investigator Sangita Gawande (middle) and her fellow officers at Nagpur Rural, having MahaCrimeOS AI as a digital companion has boosted their confidence in dealing with cybercrime. Photo by Selvaprakash Lakshmanan for Microsoft.
*The data depicted in the photo gallery is entirely fictional and is provided solely for illustrative purposes.