Trust & safety
Clinical AI that is accountable by design
Trust is earned through clear roles, visible reasoning and a complete record. These principles shape every part of the platform.
Principles
Four commitments we don't compromise on
Physicians decide
The software offers decision support only. Every finding is accepted, edited or overridden by a qualified clinician, and only a clinician can sign a report.
Transparent reasoning
When the second reader raises a flag it shows its work: the region, the overlays and a plain-language explanation, so the reader can judge it on its merits.
Full audit trail
Reads, flags, decisions and rationales are recorded in an append-only log that supports clinical governance and quality review.
Privacy by design
Data handling is designed around minimisation, access control and encryption, and is configured to meet each partner lab's own policies and obligations.
Audit trail
A complete, attributable record
Every read, flag, decision and rationale is recorded in order and attributed to a person or system — supporting governance, peer review and incident investigation.
Study received
09:02 · System
Routed from PACS to the reading worklist
Radiologist read submitted
09:15 · Radiologist
Initial impression recorded before any AI output was shown
Second reader analysis completed
09:15 · Second reader
Independent analysis of the same study
Disagreement flagged
09:16 · Second reader
Reasoning, overlays and region references attached
Radiologist decision
09:21 · Radiologist
Flag reviewed; decision and written rationale captured
Report signed
09:24 · Radiologist
Final report signed by the radiologist
Each record is append-only and attributable.
Illustrative example — not real patient data.
Practices
How the principles show up in practice
Human oversight
The workflow requires a clinician's decision for every flag and every report. The AI has no path to act on its own.
Known limitations
We document what the second reader is and isn't intended for, and share those limits with readers during onboarding.
Monitoring
Override patterns and feedback are reviewed with partner labs to understand real-world behaviour over time.
Access control
Role-based access, least-privilege defaults and attributable actions throughout the platform.
Data protection
Designed for encryption in transit and at rest, data minimisation, and retention configured to each partner's policies.
Regulatory awareness
Product availability and intended use depend on the regulatory status in each market. We discuss this openly with every partner.
Our language
How we describe what the AI does
Words matter in clinical settings. We describe the AI as a second reader that provides decision support. It raises findings for a physician's consideration; the physician interprets them and makes every clinical decision.
FAQ
Trust & safety questions
Does the AI ever communicate directly with patients?
No. Second-reader output goes to radiologists and genomics reports go to reviewing physicians. Nothing is released to patients without clinician approval.
What certifications do you hold?
We only publish certifications and regulatory clearances once they are granted and verifiable. Ask us about current status for your market during a demo.
How do you handle model errors?
The workflow assumes the AI can be wrong. Flags are presented as prompts for review, overrides are expected, and patterns in overrides are reviewed with partner labs.
Bring your governance questions
We welcome detailed questions from clinical governance, information security and procurement teams.