AI Regulation Assurance in Safety-Critical Systems

Summary

Safety-critical systems that use artificial intelligence (AI)can pose a variety of challenges and opportunities. This class of AI systems especially come with the risk of real consequential harms. With a cross-border approach spanning the UK, US, and Australia, our team aims to thoroughly investigate AI safety risks for technologies within the aerospace, maritime, and communication sectors. Through in-depth case studies, the project will identify technical and regulatory gaps and propose solutions to mitigate potential safety risks. Bridging the wider scientific community with international government stakeholders will allow us to positively impact the development and regulation of AI in safety-critical systems for the betterment of society.

Working with George Washington University in DC, and Australian National University Canberra, as well as multiple government, military, regulatory, and industry partners across the three nations.

Outputs and Outcomes

Summary of the USA workshop, September 2024

Summary report of the USA workshop held at George Washington University, Washington D.C (jointly hosted by TRAILS and RAi UK), September 2024.

This workshop consisted of four sessions and explored the broad challenges of AI trust, risk
management, safety in aviation and maritime systems, and regulatory frameworks across
international contexts. The US perspective on AI safety and assurance highlights a need for both
bottom-up (e.g., FAA’s AI Safety Roadmap) and top-down (e.g., NIST’s ARIA) approaches to
accumulate knowledge at different levels.

Download

Summary of the Australia workshop, March 2025

Summary report of the Australia workshop, held at InSpace HQ, Australian National University, Canberra, Australia in March 2025.

The workshop consisted of five sessions, including an introduction to the RAI project, goals, and strategy (presented by Jennifer Williams), followed by a summary of the US workshop held in Washington DC several months earlier (presented by Peng Wei). Throughout the sessions, we considered a technology case-study from the maritime sector on AI radio watchkeeping.

Download