90 incidents on record · 2026 Headlights Incident reports by Ellie Harris · Melbourne
10 new this week Library last updated 13 July 2026
← The incident library
HD-INC-086
Technology · United Kingdom · 2025 · AI-washing

Builder.ai sold app-building AI and raised about $445 million, then collapsed in 2025 amid reports its 'AI' was largely human engineers and its revenue far below what it had claimed

By Ellie Harris · Filed Insolvency proceedings began May 2025

Alleged: Builder.ai (formerly Engineer.ai); founder Sachin Dev Duggal developed or deployed the AI system implicated in this incident. Details are drawn from public reports; parties are presumed innocent of any wrongdoing not established by an official finding.

Builder.ai sold app-building AI and raised about $445 million, then collapsed in 2025 amid reports its 'AI' was largely human engineers and its revenue far below what it had claimed

What happened

It was reported that Builder.ai, a London-based company formerly called Engineer.ai, marketed a platform that it said could assemble most of a software application automatically, fronted by an assistant named Natasha that captured a customer’s requirements. Its founder said the system could build a large share of an app on its own. The company raised around $445 million from investors including Microsoft and the Qatar Investment Authority, reached a valuation of about $1.5 billion, and integrated with Microsoft’s Azure.

It was then reported, as the company collapsed, that the building attributed to the platform had been carried out substantially by human engineers, an estimated 700 of them, working in countries including India and Ukraine, who wrote customers’ software manually. Accounts of how far this went vary, and at least one detailed analysis has argued that the popular “faked AI with 700 engineers” description is overstated, noting that the engineers did real work and that the collapse was driven more by finances than by the human-versus-machine question. What is not in dispute is the money. In late 2024 Builder.ai borrowed from a lender, Viola Credit, which later seized about $37 million from its accounts, and in May 2025 the company entered insolvency proceedings across several jurisdictions. An independent review put its 2024 revenue near $50 million against a figure of about $220 million the company had used, and reporting described allegations of arrangements that inflated the accounts, which were the subject of further scrutiny. The founder had stepped down as chief executive earlier in 2025. Both of the claims the company was built on, how much of the work its software did and how much money it made, turned out to be far smaller than it had said.

What an auditable version would have shown

Builder.ai sold two ideas: that its software could build most of an app, and that the business was earning around $220 million a year. Both turned out to be much smaller than that. Neither had to be taken on faith. A company can record, project by project, which parts of an app its software wrote and which parts a person did. It can tie each sale to a specific invoice. An investor or an auditor can then check both for themselves instead of believing them. Builder.ai’s backers were going on what the company told them, and the real numbers only surfaced once it collapsed and someone went through the books.

Where the gap was

Both of the things that mattered here, what the software did and what the company earned, were things you had to take Builder.ai’s word for. A MetricRecord turns a number like that into a signed figure tied to the events underneath it, so an investor or an auditor can check it themselves. A ConductRecord ties each project to what the software actually did on it, so that “the AI built it” is something you can look up.

What governance should have looked like

If nobody writes down what the software did, there is no way to measure the difference between what it claims and what it delivers. The same goes for the money: a revenue figure that cannot be traced back to individual sales can be stated, but not checked. Best practice would be for a company selling an automated capability to be able to show, from signed records, how much of the work its system really does and how its revenue maps to real invoices, so that the people putting money in can check the numbers instead of trusting them.

Failure Pattern: a company’s claims about what its automated system did and what it earned were taken on its own word, and both were later put far below what it had said.

Governance Principle: a claim about what an automated system does, and the revenue attributed to it, should be verifiable from signed records rather than taken on the seller’s word.

The reference implementation of MetricRecord and ConductRecord is open source. It lives at github.com/saffronandindia/headlights-oss, Apache 2.0 licensed and free to install. The repository is public now.

Sources

The mailing list

Fresh incident reports every week. One email to match.

We add new incidents to the library regularly, and send a single short email each week with what's new. The library stays free and open; this is just how you keep up with it.

No tracking. Unsubscribe in one click.

The record

An auditable system would have produced a signed, tamper-evident record the moment this happened: what the system did, the version that did it, the basis it acted on, and the action taken, and Builder.ai (formerly Engineer.ai); founder Sachin Dev Duggal could have produced it on demand.

This is the record the system as deployed did not produce in a signed, auditable form.

What this teaches
Capture what happened when it happens
What the system did, the version that did it, the basis it acted on, and the action taken, recorded at the moment, not reconstructed after.
Sign it, so no one has to trust the record-keeper
A tamper-evident entry. Edit it later and the signature breaks. The record does not ask for the benefit of the doubt.
Make it verifiable by anyone
A court, a regulator, a customer's lawyer can check the record themselves, without taking the company, or us, at our word.

Headlights summarises publicly reported AI incidents. All summaries are independently written, attributed to their original sources, and intended for research and educational purposes. Allegations are identified as such until established through official findings.

Last reviewed June 2026. This report is based on the sources listed above and reflects information available at the time of review; later developments may not be captured. Where a person is described as charged with or alleged to have done something, that allegation is unproven unless a conviction or a court or regulatory finding is stated. Headlights publishes journalism and commentary, not legal advice.

Want to write back?

Direct to my inbox.

ellie@useheadlights.com →