The third-party committee report on alt Inc. is worth reading closely if you work in or around the technology sector. Round-trip transactions are not something you necessarily set out to join — you can be drawn into one — and simply knowing the shape of the arrangement changes how you respond.

What follows is a reading of the public report from two angles: the auditor’s, and the person actually running an accounting function.

What happened

alt Inc.’s principal product is AI GIJIROKU, a meeting transcription service. From early April 2025, during an investigation by the Securities and Exchange Surveillance Commission, the company came to recognise that revenue booked through certain sales partners might include paid accounts that were never actually used. On 25 April 2025 the board established an independent third-party committee of lawyers and CPAs to investigate the facts, the financial impact and the causes.

The committee concluded that the transactions were, in substance, round-trip transactions, and that the related revenue, advertising expense and R&D expense should not have been recognised.

What was actually circulating

The structure, per the report, ran as follows. alt recognised revenue on bulk licence sales of AI GIJIROKU to a counterparty. Separately, alt paid out funds to external advertising agencies and contractors, booked as advertising or R&D expense. Those funds found their way, through intermediaries, back to the original counterparty — which then paid alt the “sales proceeds.”

Economically, alt’s own money went out, went round, and came back. The counterparty bore almost no real cost. The committee’s framing is blunt: absent independent payment of consideration, with funds merely circulating, this is a round-trip transaction and nothing else.

The substance that should have sat behind the revenue — licences issued monthly against orders, end customers continuing to use the product — was largely absent.

The scale

The committee quantified the effect on the consolidated financial statements at approximately ¥11.9 billion of overstated revenue cumulatively across FY2020 to FY2024. For FY2022 and FY2023, the affected amount represented roughly 90% of reported revenue (91.3% and 91.0%).

That is not a misstatement at the margin. Most of the reported top line was the fraud.

The auditor’s angle: where was this catchable?

Round-tripping is a classic. It was in training material twenty years ago; there is nothing novel about the method itself. What has evolved is the concealment — from a straightforward two-party arrangement, to several same-industry parties in the chain, to this case, where advertising agencies and contractors from an entirely different sector sit in the middle.

At smaller scale these are genuinely hard to find, because the documentation is complete. Every voucher is there.

A normal audit would flag SaaS revenue recognition as higher risk and document the planned response. The report notes orders recorded in the books with little evidence that licences were ever issued or activated. But testing whether a sold licence ID shows actual usage is not something you would ordinarily build into an audit plan on resourcing grounds — absent a specific indicator of fraud.

Cash flow is the better signal. Once you can see that the source of the customer’s payment is the advertising agency or contractor, and that revenue and advertising expense are of similar magnitude, you have one of the standard round-trip indicators. Where sales and purchases are recorded against the same entity, the audit response is to treat round-tripping as a live hypothesis and design procedures accordingly.

Reporting suggests the predecessor auditor flagged a suspicion of round-tripping when handing over to the successor. If so, the hardest part of the problem — getting the initial lead — had already been solved.

From inside the accounting function

There are signals visible from the inside. A run of large deals where revenue jumps but the implementation story and usage evidence stay vague. Advertising or R&D expense rising sharply with nobody in the building who can explain the return on it. Either pattern deserves attention early.

In fairness, having worked in-house, this is realistically visible at manager level and above, where someone can see the whole shape of the transaction. Staff processing daily entries are unlikely to spot it.

The committee found internal control and governance had not functioned. The administrative and internal audit functions reported directly to the representative director, leaving little practical check, and internal audit work was insufficient. The consequence was that revenue and expense inflating in tandem — an obvious anomaly — was never surfaced as one, and nothing in the ordinary course corrected it.

In audit terms, an audit is built on the premise that internal control is effective, and the load-bearing element in that premise is management integrity. Where management is party to the fraud, the difficulty rises sharply.

This is not an exotic case

A growth story that was, in substantial part, overstatement through round-tripping — that is the committee’s picture.

From the auditor’s side, revenue existence and cash flow analysis gave a real chance of catching it. From inside the accounting function, asking whether sharply rising costs and sharply rising revenue are both matched by substance raises the odds of catching it in the daily and monthly cycle.

The wider lesson is that AI and SaaS change none of this. Remove basic accounting scepticism and governance, and the conditions for fraud assemble quickly. For anyone running a venture toward listing, pursuing the top line is not the risk in itself — failing to keep asking whether the transactions behind it are real is.