Cubictree's AI Flags ₹180 Cr Fraud Banks Missed
28 Jul 2026
A fraud that passed every check — until AI looked again
A loan worth ₹180 crore, processed through a major New Delhi institution, reportedly cleared every conventional check a bank runs before disbursing funds. It was later found to be 100% fraudulent. "It cleared all the checks of the bank. Yet it was 100% fraud," said Cubictree Founder and CEO Hitesh B. Jirawla, describing the case as an example of why the company built an AI-powered property compliance report system.
Cubictree, which has spent over a decade building this system, positions the case as evidence that AI-driven analysis can catch fraud patterns that conventional, manual compliance checks miss.
The data behind the claim
Cubictree's pitch rests on the scale of the data infrastructure it has assembled:
- 250 crore+ records in its aggregated database, with 10 lakh records added daily
- 3 to 3.5 crore unique property notices, with roughly 4 lakh new records added monthly
- 4.5 billion+ legal entries spanning over 50 years in its litigation database
- 26.5 million unique property addresses structured in its data
- Coverage of 80%+ of India's pin codes
- A client base of 100+ banks and NBFCs and 450+ enterprises across the financial and property ecosystem
Notably, nearly 75% of Cubictree's property notice records originate from physical publications — out of roughly 1,25,000 publications registered across India — rather than digital-native sources, underscoring how much of the underlying legal and property record system in India still runs on paper.
Why founders should care
The ₹180 crore fraud case, if representative, suggests that conventional compliance checks may not reliably catch sophisticated property or loan fraud — a gap that likely creates room for AI-based verification layered on top of existing bank processes. Founders building in fintech risk or lending infrastructure should probably treat data aggregation and structuring — not just algorithmic sophistication — as a meaningful differentiator, given Cubictree's scale advantage with 100+ bank clients and a decade of database-building behind it.
The heavy reliance on physical publications (75% of records) likely signals that digitization and structuring of legacy legal/property data remains a nontrivial, unsolved problem — one that could represent an opportunity for founders building complementary tools, even as it also suggests such data pipelines may be prone to delays or inconsistencies. The ten-plus years Cubictree invested before reaching enterprise scale is a useful data point for founders estimating how long comprehensive, compliance-grade data systems typically take to mature.
Reasons for caution
The report is built almost entirely around a single case study — the ₹180 crore fraud — and it's unclear how representative this is of Cubictree's broader system performance. No false positive or false negative rates were disclosed, making it difficult to independently assess the real-world reliability of the AI detection system. It's also worth noting that fraud schemes tend to evolve; a system built to catch today's patterns may face new evasion tactics tomorrow, just as this fraud evidently evaded conventional checks.
Several important details remain undisclosed: which institution processed the fraudulent loan, when it occurred, what specific mechanism the fraud exploited, and how Cubictree's AI technically differs from conventional checks. Cubictree's business model, funding history, and revenue were also not part of the available report.
The bottom line
Cubictree's scale — hundreds of crores of records, broad geographic coverage, and a substantial existing customer base — points to a real, if unquantified, opportunity in AI-driven property and litigation risk infrastructure in India. But with only one case study and no disclosed accuracy metrics, founders evaluating this space should treat the ₹180 crore fraud example as a compelling anecdote rather than proof of systematic superiority over conventional compliance checks.