One chatbot company, half a million dollars
Auckland AI firm Ambit AI was placed in liquidation on 8 June 2026 after defaulting on a $467,429 pandemic-era loan from Callaghan Innovation. The company made customer service chatbots. Grant Thornton’s Adele Hicks and Stephanie Jeffreys were appointed as joint liquidators, and creditors have until 6 July to lodge claims.
Two days after the liquidation commenced, the company changed its name to Bambi Tech. Make of that what you will.
Ambit AI’s failure is unremarkable on its own. AI startups die every week. What makes it worth examining is the programme that funded it, and the growing evidence that public capital was deployed with almost no serious follow-through.
The loan book nobody was watching
Callaghan Innovation’s Budget 2020 Covid-19 R&D loan programme pushed $150 million out to 456 firms at 3% interest over 10 years. The stated purpose was to keep R&D programmes alive when private capital dried up during lockdowns. That was defensible in 2020.
Six years later, the scorecard is grim. Sixty-seven of those 456 companies have gone bust. Only 50 have repaid in full. Nearly one-third of the $149 million loan book is in arrears, with $21.5 million linked to 63 failed or insolvent businesses. That leaves roughly 339 firms still carrying Callaghan debt, the majority of which have neither repaid nor formally defaulted. The full taxpayer exposure remains an open question.
David Farrar, writing on Breaking Views, frames it in commercial terms: “A third of the loan book in arrears would cause any other lender to go bankrupt.” His conclusion is blunt: “The Government is not good at this stuff, and should leave it alone.”
The angel investing defence falls apart
Ben Wylie-van Eerd of the Save Science Coalition offers the counterargument. He told Kiwiblog the one-third arrears rate “honestly sounds more positive than I was expecting” and compared it to angel investing: “If an angel investor had only a third of their companies that they invested in fail, they’d be jumping for joy.”
It is a fair comparison up to a point. Early-stage investing carries high failure rates by design. But Farrar’s rebuttal lands harder: angel investors deploy their own capital. They accept the risk because the upside on their winners compensates for the losers. Taxpayers get no equity upside from a 3% government loan. When the borrower fails, the public gets the downside of venture capital with none of the returns.
Labour’s science spokesperson Reuben Davidson defends the programme as emergency support that “helped companies keep their researchers employed and their programmes running when private funding dried up”. That may be true. But the emergency was six years ago, and the debt management framework apparently did not exist until MBIE inherited the mess.
MBIE is building what Callaghan never did
Callaghan Innovation was disestablished following Budget 2025, with its loan book transferred to MBIE in November 2025. The MBIE 2024/25 annual report confirms $38 million was allocated for the wind-down and transfer of Callaghan functions over three years.
MBIE’s head of innovation and business capability Diana Loughnan acknowledged the gap in oversight. Since taking over, MBIE has been “putting in place a more robust debt management framework, clearer accountability for monitoring and escalation, and a more systematic approach to identifying risks, setting impairment provisions, and recovering funds”.
Translated from bureaucrat: the previous regime lacked a proper framework for tracking who owed what and chasing them when they stopped paying. MBIE is now pursuing delinquent borrowers through the courts. Wine Grenade, another Callaghan-funded firm, was liquidated in December 2025 on application of the Attorney-General for unpaid loan obligations.
The replacement system has its own problems
The irony is that the new regime, while more fiscally disciplined, is generating fresh dysfunction. This week the NZ Herald documented a startup called Scanabull that was assessed as viable for a $750,000 commercial loan by MBIE but rejected for a smaller R&D grant. Separately, EECA’s $1 million co-investment in failed EV car-sharing firm Mevo suggests the pattern of government money flowing to startups that subsequently fail is not unique to Callaghan.
The question was never whether some pandemic loans would go bad. Of course they would. The question is whether $150 million in public capital deserved the same monitoring infrastructure a private lender would build as a matter of course. MBIE is now constructing that infrastructure retrospectively, which is an admission that Callaghan never had it. With 339 firms still carrying debt and the full impairment figure unpublished, the cleanup is far from over.
Sources
- NZ Herald: Auckland AI firm in liquidation after defaulting on $500k Callaghan loan (2026-06)
- NZ Gazette: AMBIT AI LIMITED liquidation notice (2026-06-10)
- Kiwiblog: Callaghan failure (2026-06)
- Breaking Views: David Farrar – Callaghan failure (2026-06)
- NZ Herald: Wine Grenade in liquidation – Callaghan Innovation loan unpaid
- NZ Herald: Tech Insider – Founder’s anger after MBIE denies R&D grant (2026-06-21)