550 KM · LOW EARTH ORBIT
The research
Autonomous systems already govern conduct in orbit. The rules they follow are largely private. This project examines what that means for accountability under international space law, and what a workable public framework would have to do.
01 · THE LEGAL SETTING
A regime written for human decision-makers
The treaty framework governing outer space was drafted around states, spacecraft under direct human control, and decisions taken slowly enough to be discussed. None of those assumptions describes current operations. AI systems now make thousands of irreversible decisions controlling critical space infrastructure, and they do so in a legal vacuum.
The gap is not only one of coverage. It is one of form: the obligations are procedural, requiring consultation and attribution, while the conduct is automated, distributed and fast.
OUTER SPACE TREATY 1967 · ART. IX
Requires international consultation where an activity may cause potentially harmful interference. Consultation presumes an actor able to pause.
LIABILITY CONVENTION 1972
Makes states responsible for the activities of their private actors. Attribution presumes a decision traceable to a responsible agent.
THE MISMATCH
A collision-avoidance system commits to a manoeuvre in 10 to 100 milliseconds. Diplomatic consultation runs in days.
02 · THE ARGUMENT
Private specification has become the operative law of orbit
The project’s claim is sharper than the observation that AI in space is unregulated. Satellite operators, insurers, space situational awareness providers, standards bodies and industry associations have become the effective governance actors in orbit. They set the rules through instruments that were never intended as law and are not reviewable as such.
Those instruments work. They coordinate behaviour among thousands of spacecraft under competitive conditions. But they distribute obligation according to commercial leverage rather than public standard, and they leave affected parties without recourse. The research treats this as a question of power and accountability, not of engineering capability.
TECHNICAL SPECIFICATIONS
Interface and conjunction-screening standards that determine which manoeuvre is considered correct.
INSURANCE REQUIREMENTS
Underwriting conditions that impose duties of care no statute has articulated.
DATA-SHARING AGREEMENTS
Bilateral and consortium arrangements that decide who sees a conjunction, and when.
STANDARDS PROCESSES
ISO TC20/SC14, the IADC and the Space Data Association, where norms are drafted before they are debated publicly.
03 · THE TEMPORAL PROBLEM
Decision speed as a legal fact
The obligation to consult is not merely inconvenient at machine speed; it is unperformable. The February 2024 break-up is the clearest case on record: hundreds of satellites manoeuvred autonomously within minutes, and no consultation preceded any of them. The philosophical question follows directly from the operational one.
10–100 ms
AUTONOMOUS COLLISION-AVOIDANCE DECISION
0.3 s
DECISION AFFECTING SATELLITES OF DOZENS OF NATIONS
days
CONSULTATION THROUGH DIPLOMATIC CHANNELS
The project asks not what algorithms can do, but what societies should permit algorithms to decide.
04 · APPROACH
Humanities-led, technically grounded
Legal scholarship and philosophy set the questions. Computing science establishes what is physically possible: latency budgets, sensing limits, the behaviour of autonomous systems under uncertainty. That role is deliberately delimited. What is technically feasible does not determine what is legally necessary, and the project keeps the distinction explicit.
The empirical work runs through four parallel workstreams: comparative legal analysis across seven jurisdictions, 45 stakeholder interviews, mapping of submissions to UN COPUOS and standards bodies, and synthesis into governance instruments tested with the people who would have to use them.