Onix Raises $5M for Private, Expert-Driven Health AI

The most sensitive input in health AI is often the context a person shares when they believe nobody else is listening. Onix is building its consumer product around the idea that useful personalization should not require the company to keep a central archive of those conversations.

The Montréal startup announced a $5M Pre-Seed round on September 9, 2026, led by Alpha Edison. Garage Capital, Ride Home Fund, Jeremy Zimmer, and JS Cournoyer also participated. Onix says this is its first institutional capital; independent Canadian technology publication BetaKit reports the financing was structured as a convertible note.

The money will accelerate platform development, expert onboarding, research and evaluation, and Onix's move from iOS early access toward broader commercial availability. The larger market signal is about control. Onix is trying to let experts license and govern their knowledge while users keep personal context in a device-based vault that the company says it cannot inspect.

What Happened

Onix was founded in 2025 by David S. Bennahum, Co-Founder and CEO, and Nicholas Nadeau, PhD, P.Eng., Co-Founder and CTO. The official announcement describes Bennahum as a former WIRED contributing editor and serial technology entrepreneur. Nadeau previously served as CTO of humanoid-robotics company 1X, experience that put him close to the data and system-design choices behind modern AI products.

BetaKit's independent report identifies the round as a convertible note and puts its Canadian-dollar value at about C$6.9M. The publication also reports that Onix has a 10-person team and plans to add 4 roles, mainly in go-to-market and sales. The company's current careers page shows the operating work behind that expansion: growth, product, design, expert operations, and research around small language models and expert-fidelity evaluation.

The amount is unusually large for the label. CVCA's 2025 market overview records C$121M invested across 138 Canadian Pre-Seed deals, or an average of C$0.88M. Onix's round is roughly 8 times that benchmark after the announced currency conversion. A separate CVCA seed-market analysis found U.S. participation in Canadian Pre-Seed rounds fell from 23.31% in 2024 to 8.51% in the first half of 2025, making a Los Angeles-led round in Montréal a meaningful cross-border decision.

What Onix Is Building

Onix calls its product Personal Intelligence. Instead of using one general model to answer every question from a broad public corpus, the company says each expert helps build a dedicated small language model from licensed research, frameworks, interviews, and private material. The aim is to preserve the person's methods and judgment rather than generate a generic answer dressed in a familiar voice.

That product design carries an economic choice. BetaKit reports that experts set their subscription prices, receive 70% of subscription revenue, and retain governance over how their systems operate. Onix keeps the remaining 30%. For experts, the pitch is a way to extend access without surrendering authorship or feeding another model with uncompensated work. For users, the pitch is continuity: conversations can build over time instead of restarting at each session.

Onix says personal information stays in a vault on the user's device, conversations are encrypted, and neither the company nor the expert can inspect them. The product can also use Apple Health data with permission. These are company-reported architecture and privacy claims, not an independently audited security conclusion, but they shape what Onix must deliver. A privacy promise that lives in the data flow is harder to bolt on later and easier for a customer to understand when the product behaves accordingly.

Why Health AI Raises the Stakes

Onix says more than 20 founding experts are live across health, longevity, nutrition, sleep, stress, relationships, and family care. The app is available in iOS early access. Broader commercial availability is expected later in 2026, with Android planned to follow.

The company is careful to describe the product as educational wellness guidance rather than medical advice or a replacement for clinical care. That boundary matters because a system designed to remember context and speak through a recognized expert can feel more authoritative than an ordinary chatbot. WIRED's independent product reporting found both the promise and the difficulty: participating experts valued the privacy and content controls, while early conversations also showed how an AI representation could wander outside its intended domain. One participating physician told WIRED the system would require close monitoring because AI can overstep.

That is not a side issue for a health product. Expert fidelity, escalation safeguards, consent, and clear limits are part of the product's commercial credibility. Onix must make the experience personal enough to be useful without encouraging a user to confuse continuity with clinical responsibility.

What the $5M Changes

Onix says the capital will fund platform development, expert onboarding, research and evaluation, and public availability. Alpha Edison's investment fits the firm's stated focus on early-stage companies using technology to unlock markets as behavior changes. In this case, the behavior shift is already visible: people ask general AI systems increasingly personal health questions even as trust in those systems remains fragile.

The round gives Onix room to develop a consent-based supply side and a private demand side at the same time. More experts improve coverage, but each expert also creates licensing, ingestion, evaluation, safety, and governance work. More users create richer individual context, but the architecture is supposed to keep that context outside the company's reach. Scale cannot simply mean collecting more of everything in one place.

There are no disclosed revenue, user, retention, valuation, or audited outcome figures attached to the financing. Onix's early examples and expert roster show activity, not proven consumer economics. The next evidence will come from whether people return, pay, and rely on the product within its intended limits, and whether experts believe the system represents their work faithfully enough to keep participating.

A Different AI Bargain

The dominant AI bargain has been generous distribution paired with weak provenance: enormous access to information, uncertain authorship, and user context absorbed into centralized systems. Onix is trying to make both sides more explicit. The expert supplies licensed knowledge and keeps a revenue share. The user supplies personal context and is told that context remains locally controlled.

If that structure holds through commercial launch, Onix will have built more than another wellness chatbot. It will have created a market mechanism around who gets paid for judgment, who can inspect the conversation, and where responsibility stops. The $5M now finances the unglamorous work that decides whether those promises survive contact with thousands of experts, users, devices, and deeply personal questions.