How Cubictree’s AI Protects Indian Lenders From Hidden Property Fraud Beyond Clean Title Files
Cubictree has built an AI system that cross checks property files against 4.5 billion litigation records and old newspaper notices, betting that hidden legal disputes, not forged documents, are the fraud risk Indian lenders keep missing.
Highlights:
- Cubictree runs a proprietary litigation data lake with more than 4.5 billion records, growing by 1 million cases daily
- The company’s platform helps banks assess roughly Rs 2.5 lakh crore in loan applications every year
- Cubictree cross checks property files against newspaper notices, 60 percent sourced from physical print publications
- The company currently helps manage Rs 3.8 lakh crore in default accounts across more than 100 banking systems
- Cubictree is a member of the RBI Innovation Hub and was founded by Hitesh Jirawla
A property file can look perfect on paper, every document signed, every stamp in place, every title search clean, and still be sitting on top of a legal dispute nobody in the loan file bothered to check for. That gap, between what a document proves and what it actually hides, is where Cubictree has chosen to build its entire business.
The Indian legal technology company has spent years assembling what it describes as the largest litigation data lake in the world, more than 4.5 billion records, growing by roughly 1 million new case entries every single day, and it is using that dataset to solve a fraud problem most lending institutions were never really built to catch in the first place.
Founded by Hitesh Jirawla, an entrepreneur with more than 15 years in legal technology and prior experience scaling two earlier startups alongside a stint at the Times Group, Cubictree built its core insight around a fairly uncomfortable observation about how Indian property lending traditionally assesses risk. Most fraud detection systems in mortgage lending are designed to catch forged documents, doctored income statements, fabricated identities, the kind of tampering that shows up under close forensic inspection of the paperwork itself. What they are considerably worse at catching is something quieter, a property with a perfectly legitimate looking title that is nonetheless entangled in an ongoing legal dispute, a competing ownership claim, or a default notice that never made it into any digital record a bank’s underwriting team would think to search.
Cubictree’s answer to that gap is built around a distinctly analogue insight applied at genuinely massive digital scale. A significant share of property related legal notices in India, particularly outside major metros, are still published almost exclusively in physical newspapers, in regional languages, in tier two and tier three cities that most digital risk assessment tools simply never look at. Cubictree has spent nine years building what it calls a phygital database, a hybrid repository of property notices where roughly 60 percent of entries were sourced directly from physical print publications rather than digital records, layered with AI powered search capable of working across more than 28 Indian languages.
That combination, litigation records at billion scale, paired with a searchable archive of print notices most competitors never digitised, feeds into the company’s two flagship platforms. CT-MAP, the Cubictree Mortgages Automation Platform, is built to digitise the entire property backed lending journey from onboarding through recovery, while CT-PRR, its Property Risk Report, combines litigation data with newspaper notice records to generate what the company calls a complete property risk profile before a loan is ever approved.
The scale at which banks are already relying on this system is considerable. Cubictree’s platform integrates with more than 100 banking systems, and the company says its tools currently help financial institutions manage roughly Rs 3.8 lakh crore in default accounts, while enabling risk assessment on loan applications worth approximately Rs 2.5 lakh crore annually, spanning loan sizes from Rs 40 lakh up to Rs 250 crore. Beyond risk assessment itself, the company says its infrastructure supports the issuance of 1.2 crore legal notices a year and has helped shave roughly 20 days off typical recovery timelines for lenders working through delinquent accounts.
Cubictree describes its own positioning plainly, framing its combination of litigation data and property notice records as turning what was previously scattered, largely undiscoverable information into a single, structured risk signal banks can actually act on before money changes hands, rather than discovering a legal entanglement only after a loan has already gone bad.
The company’s membership in the RBI Innovation Hub situates it within a broader, deliberate push by India’s central bank to encourage exactly this kind of data driven infrastructure inside the country’s financial system, a signal that regulators see genuine systemic value in tools that can surface property risk earlier in the lending process, rather than leaving banks to discover legal disputes only after a borrower has already stopped paying.
It is worth placing Cubictree’s pitch within the broader, considerably messier context of AI powered fraud detection in lending globally, rather than treating it as a uniquely Indian solution to a uniquely Indian problem. Mortgage and property fraud detection has become a genuine arms race everywhere, as generative AI has made it dramatically easier for bad actors to fabricate convincing documents, synthetic identities, and doctored appraisals, forcing detection tools to constantly evolve just to keep pace. Industry estimates on this broader fraud detection challenge suggest that even sophisticated AI systems currently catch consumer grade forged documents at somewhere around 85 to 90 percent accuracy, while professional grade, more sophisticated fraud attempts are caught meaningfully less reliably, a gap that widens further as fraud techniques themselves keep improving in response to each new generation of detection tools.
That broader context matters for how Cubictree’s own specific claims should be read. The company is not, on the evidence available, primarily solving the document forgery problem that dominates most conversations about AI and lending fraud, its core differentiation sits specifically in surfacing undisclosed legal entanglements and litigation history tied to a property, a genuinely different and, in India’s specific legal and administrative context, arguably under addressed risk category. Whether that specific focus makes Cubictree’s platform more or less valuable than document centric fraud detection tools depends entirely on which risk actually costs Indian lenders more money in practice, a question the company’s own published figures, focused heavily on scale of usage and recovery timelines, do not fully answer on their own.
There is also a fair question worth asking about data completeness that even a genuinely massive litigation data lake cannot entirely escape. A system built around surfacing legal disputes and print notices is only as reliable as the underlying records it can actually access, and India’s judicial and administrative record keeping, while improving, remains genuinely uneven across states, courts, and decades of accumulated paper records that have not all been digitised even by an effort as deliberate as Cubictree’s nine year phygital archive. A 4.5 billion record data lake sounds, and likely is, genuinely enormous, but enormous is not the same as complete, and the properties most likely to carry hidden legal risk, older rural land, informally transferred plots, disputes that never made it to a court that publishes accessible records, are precisely the cases most likely to sit outside even a very large dataset’s reach.
None of this undercuts the genuine value of what Cubictree has built. A clean looking property file has always been a weaker guarantee than lenders would like to believe, and a platform that can surface hidden litigation and print only legal notices before a loan closes, rather than after a default forces a bank to go looking, addresses a real and specific gap in how Indian property lending assesses risk today. Whether Cubictree’s particular combination of scale, print archive depth, and banking integration becomes the standard the rest of the industry eventually builds toward, or simply one strong entrant in a fraud detection category that will keep needing to expand its coverage as fraud itself keeps finding new blind spots to exploit, will depend less on the size of its current data lake, and more on how completely it can keep closing the gap between what a property file shows and what it is still quietly hiding.
