Glossary term
Glossary term
Governance and Compliance
Structured argument that a system is acceptably safe for a defined use.
The UK MHRA requires AI software as a medical device (SaMD) to include a safety case in submissions - Babylon Health's triage AI included a structured safety argument evidencing clinical accuracy and escalation coverage.
NASA's software safety standard NPR 7150.2 requires safety cases for flight-critical software - the pattern is now being adopted by AI governance teams for high-risk agent deployments in critical infrastructure.
A global bank's AI governance team produces a safety case for each Tier-1 AI system before go-live: hazard identification, risk controls, residual risk acceptance, and post-deployment monitoring commitments.