63,000 Restaurants. One Software Company. And the AI That Caught Them All.
How a biryani shop in Hyderabad accidentally exposed the biggest tax scandal in India’s restaurant history
63,000 Restaurants. One Software Company. And the AI That Caught Them All.
How a biryani shop in Hyderabad accidentally exposed the biggest tax scandal in India’s restaurant history

There is a button in most restaurant billing software that nobody talks about.
It lets you go back into a closed transaction and delete it. Quietly. Without a trace. The bill existed, the customer paid, the money came in and then, with a few clicks, it never happened.
For six years, tens of thousands of restaurants across India pressed that button.
Last month, the government found out.
It started small.
November 2025. Hyderabad. A handful of income tax officers walk into a popular biryani chain for what should be a routine check. They pull the billing records. They compare them to the filed returns. Something doesn’t add up, not by a little, but by a lot.
They dig deeper. They find the deletions. And then they realize something that changes everything: this isn’t one restaurant. This isn’t one chain. This is a feature of the software itself.

The same billing platform used by over one lakh restaurants across India had, buried in its backend, the ability to edit and delete transactions after they were recorded. Restaurants weren’t hacking anything. They weren’t breaking into systems. They were using a function the software gave them and using it to make revenue disappear.
What happened next was unlike anything India’s tax authorities had done before.
They didn’t send officers to restaurants. They went to the source.
The Income Tax Department obtained access to the billing software company’s servers in Ahmedabad. They copied everything. Sixty terabytes of raw transactional data. Six financial years. ₹2.43 lakh crore worth of billing records across 1.77 lakh restaurant identification numbers.

Then they did something the restaurants never anticipated. They loaded it all into high capacity computing systems and ran generative AI across it.
The AI didn’t need to understand restaurants. It just needed to find patterns. And the pattern was everywhere.
Transactions recorded. Then deleted. Not randomly, selectively. Cash entries, specifically. At the end of days. At the end of months. Right before tax filing deadlines.
The AI flagged them all. Every deletion. Every timestamp. Every GST number attached to a restaurant that had billed ₹10 lakh but declared ₹6 lakh.
What took six years to accumulate took weeks to unravel.
The numbers landed like a verdict.
₹13,317 crore in post-billing deletions flagged across the dataset.
₹70,000 crore in estimated concealed turnover since 2019.
25 to 27% of restaurant sales across India potentially suppressed through digital manipulation.
On March 8, 2026, the CBDT conducted a single coordinated nationwide survey. One day. 62 restaurants. 46 cities. 22 states. Sales suppression found in just those 62 outlets: ₹408 crore.
Karnataka led the state-wise breakdown at ₹2,000 crore in deleted transactions. Telangana at ₹1,500 crore. Tamil Nadu at ₹1,200 crore. Maharashtra and Gujarat close behind.

63,000 restaurants received advisory messages from the Income Tax Department in the days that followed.
Here is what most coverage of this story gets wrong.
The headlines called it a scam. A scandal. A racket. And for many of the restaurants involved, it was exactly that deliberate, systematic, calculated evasion.
But a significant number of restaurants receiving those 63,000 advisories didn’t think of themselves as criminals. They thought of themselves as survivors.
Margins in the Indian restaurant industry run at 3 to 6% for most formats. Rent is fixed. Labour costs are rising. Aggregator platforms take 25 to 30% of every delivery order. GST on restaurant services runs at 5% for most formats but the compliance burden is real and constant.

In that environment, the delete button in the billing software stopped feeling like fraud and started feeling like arithmetic.
That doesn’t make it legal. But it does explain why the practice spread so widely, so quietly, across so many cities and formats and cuisine types. It wasn’t organised crime. It was an industry coping mechanism, one that the government just decided, with considerable computing power, to stop tolerating.
The window is closing.
The CBDT launched the SAKSHAM NUDGE campaign alongside the enforcement action, a deliberate choice to lead with amnesty before punishment. Restaurants that received advisories can file updated returns under Section 139(8A) of the Income Tax Act before March 31, 2026. Pay the tax. Pay the penalty. And the matter is treated as a compliance correction rather than a criminal case.
After March 31, that changes.
The department has signalled it will move to stricter enforcement against non-compliant entities. More billing platforms will come under scrutiny, the Ahmedabad company is explicitly described as one of several that operate in the sector. The investigation is widening, not concluding.
For the 63,000 restaurants that got a message: the deadline is real.
For the ones that didn’t: the investigation isn’t finished.
The deeper story here isn’t about tax evasion.
It’s about what happens when an entire industry builds its operations around the assumption that certain things will never be checked.
Cash has always been the soft underbelly of the restaurant business. It moves fast, it leaves no trail by default, and in a high-volume environment it becomes very easy to blur the line between what came in and what gets declared. The billing software deletion trick worked for six years not because it was clever but because nobody was looking at the right data in the right way.
AI changed that. Not gradually, but overnight.
The government now has a capability it didn’t have two years ago: the ability to ingest terabytes of transactional data, cross-reference it against GST filings, PAN records, and open-source restaurant databases, and surface discrepancies at scale without a single human auditor reading a single spreadsheet.
That capability is not going away. It will only get cheaper, faster, and more comprehensive.
The restaurant that built its operations around cash flexibility has a problem that isn’t solved by filing an updated return. The problem is structural. Every cash transaction that isn’t fully reflected in the books is now a potential data point in a system that is actively looking for exactly that pattern.
So what does a restaurant actually do with this?
The honest answer is: move toward digital, systematically, and not just for compliance reasons.
Every UPI transaction is immutable. Every card payment has a timestamp and a record. Every QR order generates a digital trail that matches the payment to the item to the time to the table. There is no delete button for a UPI transaction, the money moved, the record exists, the trail is complete.
Restaurants that have already shifted toward digital ordering and digital payments are not just ahead on customer experience. They are structurally insulated from exactly the kind of investigation that is currently sweeping the industry.
Not because they’re doing anything special. Because their operations, by design, leave nothing ambiguous.
The March 31 deadline will pass.
Some restaurants will file. Others will hope the government is bluffing. It isn’t.
More software platforms will be examined. More data will be loaded into more machines. More patterns will be found.
This isn’t a story about 63,000 restaurants that got caught. It’s a story about an industry at an inflection point, one where the infrastructure you choose to run your business on will determine whether you’re protected or exposed when the next wave of scrutiny arrives.
The bill always existed. Now it always will.
If you run a restaurant in India and want every order, payment, and transaction recorded with a complete digital trail, Menuthere builds that into your operations from day one.
메타데이터
- post_id
- 862d63afdace
- slug
- 63-000-restaurants-one-software-company-and-the-ai-that-caught-them-all-862d63afdace
- url
- https://medium.com/@menuthere/63-000-restaurants-one-software-company-and-the-ai-that-caught-them-all-862d63afdace
- canonical_url
- https://medium.com/@menuthere/63-000-restaurants-one-software-company-and-the-ai-that-caught-them-all-862d63afdace
- author_url
- https://medium.com/@menuthere
- status
- ok
- fetched_at
- 2026-06-28 10:39:35