Halcyon Arc / Guardian platforms / AI Governance Platform
Guardian platform · Built in-house

Every AI decision, accountable to a person.

The Halcyon Arc AI Governance Platform is our system of record for responsible AI: which models we run, why we trust them, who approved them, what data and consent they rely on, and what happens when something goes wrong.

MODEL REGISTRY · EVERY MODEL OWNED & APPROVED
Why we built it

Health AI has to be explainable long after launch.

Patients, families, clinicians and regulators will ask hard questions: why did the system say that, who decided it was ready, was my data used, and what changed since last month? Spreadsheets and scattered documents can't answer them reliably. The AI Governance Platform keeps the answers in one place, linked to the evidence.

  • A complete inventory of models, prompts and data sources
  • Release gates that require evidence and a named approver
  • Consent and audit records that hold up under review
MODEL REGISTRY
Care companion v0.4
Approved
RISK REGISTER
3 open · 0 high
Tracked
CONSENT LEDGER
Detection research cohort
Current
INCIDENTS · 90 DAYS
Patient-impacting events
0
Illustrative governance console.
Capabilities

The record behind every release

AI inventory and model cards

Every model, prompt, dataset and third-party service is registered with an owner, intended use, limitations and current status.

Risk and impact assessments

Structured assessments for each product and major change, including algorithmic impact assessment and clinical risk review.

Release gates and sign-off

Nothing ships until AI Safety evidence is attached and a named clinical or product lead approves it.

Consent and data lineage

Trace which data feeds which model and under what consent. When someone withdraws consent, it flows through to every use.

Immutable audit trail

Tamper-evident logs of approvals, access to health information and model changes, ready for privacy and regulatory review.

Incidents and change control

A clear path to report, triage and resolve AI incidents, roll back a model and learn from every post-incident review.

How it works

Register, assess, approve, oversee

Register

Each model and data source enters the registry with an owner, purpose and consent basis.

Assess

Risks and impacts are assessed with clinicians, privacy and security, and logged with mitigations.

Approve

AI Safety results are attached as evidence and a named approver signs off the release.

Oversee

Monitoring, incidents and periodic reviews keep every record current for as long as the model runs.

Frameworks we align with

Governance mapped to the rules health care runs on

Our records are structured so they map directly to management-system, privacy and medical-device expectations.

ISO/IEC 42001 AI management systemNIST AI RMF (Govern function)PIPEDAOntario PHIPAHealth Canada software as a medical device guidanceAlgorithmic impact assessmentModel cards & datasheets
The Halcyon Arc family

Healthcare products, guarded by design

Our dementia products are the first undertakings. The AI Safety and AI Governance platforms are the guardians we built to stand behind each of them.