The Office of Rail and Road (ORR), the UK’s rail and road regulatory authority, has published an action plan for the safe adoption of artificial intelligence in the rail sector and on England’s strategic road network. The document outlines how the ORR will clarify safety risks, authorization processes, data usage, and potential controlled trials for AI-based technologies.
The plan, titled Safe AI Innovation Action Plan 2026, responds to a request sent to the ORR by UK ministers responsible for science, innovation, business, and transport. The aim is to demonstrate how AI-based innovation can be supported without compromising safety and regulatory obligations.
ORR states that its role is not to regulate artificial intelligence as a technology in itself, but to ensure that the use of AI in regulated activities remains transparent, governable, and auditable and compatible with existing legal obligations. The Authority states that it will apply current safety and economic frameworks in a proportionate and results-oriented manner, without introducing new rules solely for a specific technology.
AI in Railway Digital Safety
One of the ORR’s first steps will be to publish a digital safety strategy and a strategic risk chapter on the health and safety risks arising from the use of digital systems. These documents are to be finalized in the second quarter of the 2026/2027 fiscal year.
Through this strategy, the ORR will establish how digital risks, including those associated with AI, should be integrated into the safety management systems of operators and infrastructure managers. The Authority emphasizes that these risks must be treated the same as traditional safety risks, not as a separate or experimental category.
The documents will clarify the ORR’s expectations regarding the management of digital risks and will emphasize cooperation between manufacturers and railway companies when new digital systems are introduced. The aim is to reduce regulatory uncertainty so that the industry can invest in innovative solutions with greater clarity.
Existing rules applied to new technologies
The ORR will also update the guidance on applying the legal obligations in the ROGS, the UK’s railway safety regulatory framework, as well as in the Common Safety Method for risk assessment.
The update will explain how these obligations apply to the development and use of digital technologies, including AI.
The authority found, following a post-implementation assessment of ROGS, that there is a low level of awareness within the industry of the resources already available, including standards and guidance. Therefore, ORR wants to explain more clearly how the existing legal framework can be used for new technologies, without creating a parallel system of rules.
This approach is important for the rail sector, where AI can be used in areas such as maintenance data analysis, infrastructure inspection, traffic planning, rolling stock authorization, or monitoring of passenger services.
AI in train and infrastructure authorization
The ORR’s plan also includes an analysis of how AI can be used in the interoperability authorization process. This process verifies whether new, modernized, or refurbished rail vehicles and infrastructure comply with common technical standards and can operate safely on the main network.
ORR will analyze whether AI tools can support faster and higher-quality assessment of authorization applications. At the same time, the authority will also examine the use of AI by assessment bodies or by applicants, including for generating evidence of compliance with technical standards.
For the railway industry, this aspect is relevant because the authorization of vehicles and infrastructure can be a complex process involving large volumes of technical documentation. However, ORR notes that any use of AI must be evaluated within existing processes, not treated as a shortcut that eliminates human verification.
Complaint analysis and passenger protection
The ORR also intends to use AI-based analytical tools to monitor passenger-oriented services. The plan explicitly mentions complaints, passenger assistance, and datasets from surveys.
The Authority wants to see if AI can more quickly identify recurring issues, emerging risks, or causes of passenger dissatisfaction that are not visible through standard manual analysis. These tools could help ORR intervene earlier in the event of issues regarding service quality or accessibility.
At the same time, ORR notes that data will play an important role in the decision-making process, but without reliance on sensitive or identifiable data. The plan calls for exploring the possibility of using anonymized or synthetic datasets for the development and adoption of AI.
Open data and controlled tests
The ORR says it will work with data providers to understand the barriers limiting the publication of information and to identify areas where anonymized or synthetic datasets can fill existing gaps. The authority also mentions the possibility of making data available through the Rail Data Marketplace, the UK platform for rail data exchange, or through a new ORR data portal application.
The plan also includes the possibility of controlled testing, such as a regulatory sandbox, where AI adoption is slowed by uncertainties regarding the regulatory framework. ORR will analyze use cases in the rail sector, including through the priority initiatives of GBRX, the strategic technology body for the British railways.
Areas mentioned include infrastructure asset management, automated train scheduling, and network operations. In some cases, testing could take place within a structured, time-limited framework, with appropriate safeguards and oversight.
ORR is already using AI internally
The Authority has invested in digital and data infrastructure, including analytics platforms and a modern data storage and analytics architecture, to support advanced analytics and future AI applications.
One example mentioned is the Workshopping AI Support Individual (WAISI) tool, used to support discussions within ORR’s asset management team. The tool quickly searches for relevant evidence in large and complex datasets and is designed to show the steps taken, so that users can validate the sources and reasoning.
For ORR, this is part of a broader strategy to build internal capacity to use AI in a controlled manner, with data protection policies, mandatory staff training, and risk management mechanisms.
Six actions by 2027
The ORR plan includes six key actions: publishing the digital safety strategy, updating guidance on legal obligations, analyzing the use of AI in the interoperability authorization process, using AI for passenger-oriented monitoring and analysis, evaluating anonymized or synthetic datasets, and reporting on innovative methods, including regulatory sandboxes.
Stephanie Tobyn, Director of Strategy, Policy, and Regulation at ORR, stated that the plan demonstrates how the authority is contributing to the responsible adoption of AI, with users at the center of innovation.
“We are providing regulatory stability to the industry regarding what we expect from it and, at the same time, offering users the assurance that, in the industries we regulate, AI can be used responsibly, transparently, and safely,” said Stephanie Tobyn.
For the British rail sector, the plan marks a shift from general discussions about AI to defining concrete steps: clarifying risks, adapting authorization processes, using data, and conducting controlled testing of use cases. In a system already undergoing reform, with the launch of Great British Railways, AI is becoming part of how British authorities are seeking to modernize both the regulation and operation of the network.
Share on:
