DNV Publishes Blueprint for Safe Industrial AI — Maritime and Energy Experience as Foundation

DNV published new research on March 25, 2026, identifying the foundations required to achieve trustworthy artificial intelligence in safety-critical industrial processes — drawing directly on decades of assurance experience from the maritime and energy sectors to establish a practical framework for AI governance as the technology embeds deeper into vessel and infrastructure operations. Christian Agrell, Programme Director for AI Assurance at DNV, stated that the building blocks for safe AI are already in place, and the challenge now is applying them systematically as AI moves from isolated applications to integrated operational systems. Core framework principles: At the centre of DNV's approach is a system-wide view of AI rather than focusing on individual algorithms or models in isolation. The research stresses that companies need to understand how AI interacts with people, physical assets, and digital systems — and how risks emerge from those interactions rather than from the AI component alone. The framework calls for modular risk models to break down complex systems into assessable components, alongside clear links between safety claims and verifiable evidence. This creates a more transparent and auditable pathway to demonstrating that AI systems remain safe over time. Continuous assurance requirement: Critically, the report underlines that assurance must be ongoing — not a one-time certification. As AI models evolve, data shifts, and operating conditions change, companies need to monitor performance in real time and regularly reassess risks. This represents a fundamental departure from traditional maritime certification models where class certificates typically have three-to-five-year validity periods. Why this matters for ship managers: For operators deploying AI in navigation, maintenance, or compliance systems, DNV's framework signals the direction of future regulatory requirements. The message is clear: AI deployed in safety-critical maritime operations will require continuous monitoring, transparent evidence chains, and system-level risk assessment — not just a certificate at installation. Ship managers should begin evaluating whether their AI vendors can demonstrate the kind of ongoing assurance DNV describes, as these standards are likely to become classification requirements. [Splash247]

This story is part of the Maritime AI Digest of 29 March 2026.