T022 // Space Law and Artificial Intelligence (Part 3)

Artificial Intelligence (AI) is already beginning to proliferate throughout the space industry. AI is being used to process data captured in space for the benefit of life on Earth. It is also being used in outer space itself to help with missions on board satellites, rockets, and rovers, enabling a new generation of space exploration capabilities. As AI capabilities spread and develop further, they will increasingly come into contact with the international space law framework and the interaction between AI and space law will become more relevant and need to be properly understood. So how does AI interact with the international space law regime?

The core of international space law (the treaties) was drafted in the 60s and 70s, well before modern AI came into existence. Because of this, AI is not addressed directly in any of the international treaties governing space activities. This doesn't mean they are irrelevant though! They still form a foundational basis for regulating the use of AI in space and will be relevant for a long time to come.

The Outer Space Treaty of 1967 set out the key principles of space law. Some of these principles, including the responsibility of States for national activities in outer space, the liability of States to other State parties for damages caused by their space objects, and the jurisdiction of States over objects sent into space. The Liability Convention then expanded the liability regime, and the Registration Convention detailed the rules regarding how space objects should be registered and what information should be recorded. Beyond treaties, other documents may be relevant, but there are no sources of guidance directly addressing AI or automated systems.

The lack of guidance isn’t outright problematic, but it does raise some questions. For instance, what happens if an AI system is ‘autopiloting’ a satellite in orbit, and it accidentally crashes it into another satellite, destroying them both? Fault would need to be established in order to hold the AI enabled satellite’s operator liable. But given there is already a lack of ‘case law’ in space, with no solid examples of the treaties being applied to real scenarios, what happens when you add the complication of an opaque AI system to the mix? How can you tell exactly what went wrong? Was there actually ‘fault’, or was the AI system’s sensors fooled like a Tesla mistaking the full moon for a yellow traffic light? Where does the fault lie in the chain of interconnected systems that the AI component operates within? Does an AI system falling short of a specified standard of reliability classify as fault? For now, while academics may debate these questions, there are no solid solutions.

Those academics haven’t been resting though. Many different propositions have been advanced to try and address these questions. New treaties and binding protocols to existing treaties could clarify details of the existing framework to make them ‘future ready’. A ‘special registry’ for non-tangible space objects could track complicated software systems being used in space to help with information sharing and potentially liability claims in the event of an accident. However, it is unlikely these suggestions will be adopted anytime soon, and so it’s more likely that the space industry will have to look to AI specific regulations from non-space legal frameworks for guidance.

Indeed, the landmark adoption of the European AI Act demonstrates this commitment, alongside initiatives like the American Executive Order on the Safe, Secure and Trustworthy AI and the UN Resolution on Responsible AI. The international competition to shape the trajectory of AI governance is booming. In future editions, we’ll explore the different AI related initiatives that could inform the use of AI in the space industry!