Healthcare AI, with guardians built in.
Halcyon Arc is a Canadian healthcare AI company. Our first undertakings are dementia detection and dementia care: tools that help families and clinicians notice change sooner and make every day at home a little calmer. Every product runs under two guardian platforms we built for ourselves, AI Safety and AI Governance.
Dementia is our first undertaking, not our last.
Nearly every family is touched by dementia, and most of the care happens at home with little support. We start here because the need is large, the moments are human, and the bar for trustworthy AI is as high as it gets.
Halcyon Arc forDementia Detection
Earlier, gentler screening for cognitive change using short speech, language and memory tasks at home.
Explore Dementia Detection Healthcare productHalcyon Arc forDementia Care
A calm, voice-first companion for people living with dementia, their caregivers and care teams.
Explore Dementia CareNext for Halcyon ArcHealth guidance & mental health
Dementia is where we start. We plan to bring the same guarded approach to everyday health questions and mental-health support.
Our directionTwo platforms stand between our AI and the people it serves.
We built the AI Safety and AI Governance platforms for ourselves, because no off-the-shelf tool met the bar that health care demands. Every Halcyon Arc product has to pass through both.
Halcyon Arc AI Safety Platform
Clinical evaluation suites, red-teaming for vulnerable users, fairness testing across accents, languages and ages, and runtime guardrails that screen every live conversation for unsafe advice, prompt injection and personal-data leaks.
Halcyon Arc AI Governance Platform
A registry of every model we run, risk and impact assessments, release gates with clinical sign-off, a consent and data-lineage ledger, immutable audit trails and an incident process, so every AI decision is accountable to a person.
Nothing reaches a patient or caregiver without passing both.
The people who most need AI's help are often the most vulnerable to its failures. A memory-care companion that can be confused or manipulated, or a screening model that leaks voice recordings, causes real harm. So safety and governance are not features we add at the end. They are the gates every release passes through.
- AI Safety evaluates every model before release and screens every interaction after it.
- AI Governance records who approved each release, on what evidence and under which consent.
- Clinicians and caregivers stay in charge of every consequential decision.
Calm, careful and accountable
Safety before scale
We release only what our AI Safety Platform can evaluate, explain and roll back. Red-teaming is part of shipping, not an afterthought.
Canadian by default
Health data is hosted in Canadian cloud regions and handled in line with PIPEDA and provincial health-privacy law such as Ontario's PHIPA.
People stay in charge
Our AI informs clinicians, caregivers and families. Consequential decisions about health remain with people.