35,786 KM · GEOSTATIONARY
Governing AI in the Final Frontier
A Multi-Stakeholder Approach to Autonomous Space Systems Regulation
AI systems now make thousands of irreversible decisions controlling critical space infrastructure. They do so in a legal vacuum — and the rules that do bind them are being written by satellite operators, insurers and standards bodies rather than by states.
20,000 KM · THE CRISIS
In February 2024 a decommissioned Russian satellite broke up in orbit. Astronauts aboard the ISS sheltered in place. Within minutes, AI aboard hundreds of satellites had autonomously executed avoidance manoeuvres.
~50,000
autonomous collision-avoidance manoeuvres performed by Starlink between December 2023 and May 2024
~15,000
satellites expected to launch in the next five years
0.3 s
the span in which an autonomous decision is taken affecting satellites from dozens of nations
SOURCES · SPACEX, SEMIANNUAL CONSTELLATION STATUS REPORT TO THE FCC, 1 DEC 2023 – 31 MAY 2024 · LAUNCH FORECAST AND DECISION-WINDOW FIGURE: CITATION TO BE ADDED
SOURCE · EUROPEAN SPACE AGENCY, SPACE DEBRIS OFFICE
550 KM · LOW EARTH ORBIT · THE ARGUMENT
Non-state actors have become the real governance actors in orbit.
Satellite operators, insurers, space situational awareness providers, standards bodies and industry associations create de facto rules through technical specifications, insurance requirements and data-sharing agreements. These instruments regulate conduct in orbit effectively, continuously, and largely outside treaty frameworks.
The question this project puts is therefore one of power and accountability rather than technology. Where private specification has taken the place of public law, who is answerable for the decisions it authorises — and by what standard?
100 KM · KÁRMÁN LINE · THE TEMPORAL PROBLEM
Space law assumes agents who can pause and consult.
Article IX of the Outer Space Treaty requires international consultation where harmful interference may occur. The Liability Convention holds states responsible for their private actors. Both presume deliberation. An autonomous collision-avoidance system decides in 10 to 100 milliseconds and is physically unable to wait for a process that runs in days through diplomatic channels.
The project asks not what algorithms can do, but what societies should permit algorithms to decide.
MACHINE DECISION
10–100 ms
TREATY CONSULTATION
days
CONSULTATION RUNS ROUGHLY ONE MILLION TIMES LONGER THAN THE DECISION IT IS MEANT TO PRECEDE
50 KM · FOUR PARALLEL WORKSTREAMS
ALL FOUR IN DETAIL →WS1
Comparative Legal Analysis
Seven jurisdictions, and how the 1967 and 1972 treaties meet private governance.
WS2
Stakeholder Engagement
45 interviews with operators, insurers, SSA providers and standards bodies, plus co-design.
WS3
Power Dynamics Mapping
Submissions to UN COPUOS, ISO TC20/SC14, the IADC and the Space Data Association.
WS4
Framework Integration
Synthesis into governance tools, validated with regulators and industry.
The project is humanities-led. Technical colleagues hold a deliberately delimited role: they establish what is physically possible for legal scholars working on what is legally necessary.
10 KM · EIGHT INSTITUTIONS · SIX COUNTRIES
THE CONSORTIUM →LEAD · UNITED KINGDOM
University of Aberdeen
UNITED KINGDOM
University of Liverpool
AUSTRALIA
Australian National University
CANADA
University of Montreal
GREECE
National and Kapodistrian University of Athens
INDIA
O.P. Jindal Global University
GHANA
University of Cape Coast
INDUSTRY PARTNER
NIPARO
1 KM · WHAT THE PROJECT PRODUCES
Programme
Four phases across 24 months, from foundation and mapping through to implementation with regulators.
2026–2028 →Outputs
Comparative legal reports, an audit-ready accountability framework, model clauses, an interactive toolkit and open-access briefs.
ALL OUTPUTS →Events
Co-design sprints, roundtables in London and Montreal, and presentations at the International Astronautical Congress.
PROGRAMME OF EVENTS →Team
Ten researchers in space law, AI ethics, information governance and regulatory frameworks, with technical grounding.
WHO IS INVOLVED →Toolkit
The framework, typology and model clauses in a form a regulator or operator can work through directly. In development.
WHAT IT WILL CONTAIN →Participate
Operators, insurers, SSA providers, standards bodies and regulators are invited to take part in the interviews and co-design sessions.
HOW TO TAKE PART →