Kintsugi Shuts Down After 7 Years, Open-Sources Depression AI
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Kintsugi shut down after seven years of developing speech-based depression and anxiety detection AI, releasing most of its technology as open-source after failing to secure FDA clearance and burning through its funding
- CEO Grace Chang said the FDA's "De Novo" pathway — designed for novel, low-risk medical devices — is poorly suited to AI because it treats models like fixed devices such as hip implants or pacemakers, and the process was further delayed by federal government shutdowns
- Kintsugi rejected what Chang called a "predatory" offer of roughly $50,000 per week in exchange for $1 million in equity rather than accept short-term funding that would have given investors favorable terms
- Kintsugi is withholding its deepfake voice-detection capability from the open-source release for security reasons, a feature the team discovered incidentally when AI-generated speech used to train mental health models turned out to be distinguishable from human voices
- Nicholas Cummins, a senior lecturer in speech analysis and responsible AI in health at King's College London, warned that open-source releases typically lack the regulatory "paper trail" required for FDA approval, and cautioned that voice-based mental health systems carry a "reasonable" risk of errors across different languages and cultural contexts
- Chang dismissed misuse concerns as less practical than they appear in theory, arguing the organizations most likely to abuse the technology also face the highest barriers to deploying it — "the more realistic risk is underuse, not misuse"
Why it matters: Kintsugi's shutdown shows what happens when an FDA pathway built for static devices like hip implants meets a fast-moving AI company: seven years of work, rejected predatory funding, and open-sourced code that experts warn could be deployed by employers or insurers without clinical safeguards — a cautionary structure for every AI health startup racing the same regulatory clock.
Ask SkimNews
