DNV Introduces Framework for Trustworthy AI in Safety-Critical Industries Based on Proven Risk Management Principles

DNV Introduces Framework for Trustworthy AI in Safety-Critical Industries Based on Proven Risk Management Principles

(IN BRIEF) DNV has published new research identifying key principles for building trustworthy artificial intelligence in safety-critical industrial settings, emphasizing that established risk management and assurance practices can be adapted to meet the challenges of AI-enabled systems. The report highlights the need for continuous, adaptive assurance rather than traditional one-time evaluations, given the evolving and unpredictable nature of AI technologies. Drawing on experience from industries such as maritime and energy, DNV outlines foundational approaches including system-level modelling, modular risk assessment, evidence-based validation and ongoing monitoring throughout the lifecycle. The findings provide practical guidance for organizations aiming to safely deploy AI while maintaining reliability, transparency and accountability in complex operational environments.

(PRESS RELEASE) LONDON, 25-Mar-2026 — /EuropaWire/ — DNV has released new research outlining key principles for achieving trustworthy artificial intelligence in safety-critical industrial environments, highlighting how established risk management approaches can be adapted to address the unique challenges posed by AI-enabled systems. The report, titled Assurance of AI-Enabled Systems, emphasizes that while artificial intelligence introduces new layers of complexity and uncertainty, existing assurance frameworks from high-risk industries provide a solid foundation for managing these risks effectively.

The study explains that AI systems differ fundamentally from traditional technologies, as they do not behave as fixed and fully predictable components. This dynamic nature means that conventional, one-time assurance methods are no longer sufficient. Instead, organizations must adopt continuous and adaptive assurance practices that evolve alongside the systems themselves, ensuring safety and reliability over time.

Drawing on decades of experience in sectors such as maritime and energy, DNV identifies several core principles necessary to support the development and operation of trustworthy AI. One of the central elements is the creation of a comprehensive system model that captures the full scope of AI-enabled systems. This approach considers how AI interacts with human operators, digital infrastructure and physical assets, enabling a deeper understanding of complex behaviors, unintended consequences and context-specific risks that may not be visible when evaluating AI components in isolation.

The research also highlights the importance of a modular approach to risk management. By applying uncertainty-based assessments and breaking down complex systems into manageable components, organizations can better identify and control risks at different levels of the system. This structured methodology helps address the emergent and interconnected risks associated with AI technologies.

Another critical aspect involves linking safety and performance claims to verifiable evidence. By establishing transparent and auditable arguments that connect system claims to supporting data, assumptions and reasoning, organizations can demonstrate trustworthiness in a consistent and structured manner throughout the lifecycle of AI systems.

The report further underscores the need for ongoing, context-aware assurance processes. As AI models evolve, data inputs shift and operating environments change, maintaining trust requires continuous monitoring, regular updates to supporting evidence and periodic reassessment of risks. This lifecycle-based approach ensures that systems remain safe, reliable and aligned with operational requirements over time.

DNV positions these principles as practical guidance for industries seeking to implement AI responsibly, particularly in environments where safety and reliability are critical. The research forms part of the company’s broader efforts to support the safe adoption of AI and aligns with its recommended practice for AI assurance, providing organizations with a structured pathway to manage both the opportunities and risks associated with advanced technologies.

Peter Lovegrove
Media Relations Manager, Group
+47 409 04 294

SOURCE: DNV