Norwegian regulator flags AI risk management issues at Yggdrasil project

Regulatory scrutiny of offshore AI

The Norwegian Ocean Industry Authority, known as Havtil, is increasing its oversight of how energy companies integrate artificial intelligence into high-risk environments. As offshore operators move toward more autonomous facilities, the regulator is conducting specific audits to ensure these technologies do not compromise safety. The focus is currently on how companies qualify new tech and manage the risks associated with it.

Aker BP's Yggdrasil development in the North Sea has become a primary focus for these audits. The project is considered one of the industry's most advanced digital fields, making it a high-profile test case for large-scale AI deployment. Because the project is so visible, the regulator's findings will likely set a precedent for how other companies approach similar automation goals.

The audit specifically looked at the requirements and systems Aker BP has put in place to ensure safe operations. Havtil wants to confirm that the company understands the nuances of AI-assisted decision-making. The regulator noted that technology qualification—the process of proving a system is fit for its intended purpose—remains a significant hurdle for the industry.

The automation at Yggdrasil

The Yggdrasil project is designed to be highly automated, with much of the control and monitoring conducted from onshore locations. This model aims to reduce the number of people physically present at offshore production facilities, which can sometimes have limited or no staff. By moving operations to the shore, companies can improve efficiency and reduce the risks associated with human presence in harsh environments.

Aker BP plans to use a variety of specific technologies to achieve this goal. This includes remote operations and robotic inspection tools that can detect gas leaks, heat generation, and other abnormal conditions. Some of these robots will gradually take over routine inspection duties that were previously performed by humans on-site.

The company also intends to install fully electric cranes that can be operated from the shore. In these scenarios, AI models will provide real-time information to the crane operator. The goal is to help the human operator detect and manage potential incidents more effectively. However, the regulator found that the current plans for these robotic deployments do not meet specific safety standards.

Identified non-conformities

During the audit, Havtil identified two specific non-conformities related to the planned use of robots at Yggdrasil. These issues are tied to technology qualification and the company's understanding of risk management. Essentially, the regulator found that the current framework for approving and monitoring these autonomous systems is not sufficient for the risks involved.

These findings come as the industry faces a broader push toward autonomous systems. Havtil has already signaled that many companies in the petroleum sector need to improve their oversight and competence regarding AI. The regulator's focus is on ensuring that these tools are properly understood before they are allowed to perform critical tasks in offshore drilling or production.

Aker BP is now required to provide a formal response to these concerns. The company must submit a planned action to address the regulator's findings by November 15. This move highlights the necessity for energy companies to develop more robust evidence-based systems to validate the safety of their automated tools.

The path toward autonomy

Industry experts suggest that while AI can provide significant value, it requires human accountability. For example, AI can help prioritize maintenance tasks and monitor safety data, but engineers must still validate the evidence provided by the software. This is especially important for life-extension decisions where the consequences of a mistake could be severe.

The use of edge analytics and thermal imaging can strengthen safety monitoring and help target maintenance more accurately. These technologies can also reduce the number of unnecessary offshore visits, which is a primary goal for many operators. By using AI to identify specific problems, companies can send teams only when and where they are truly needed.

Moving forward, the Yggdrasil project will serve as a benchmark for how these technologies are integrated into the North Sea. The regulator's actions suggest that the era of 'deploy and see' is over for industrial AI. Companies will need to demonstrate a deep understanding of their systems' limitations and capabilities before they can achieve full autonomy.