Autonomous AI inference in zero gravity operates beyond human oversight. No communication link. No review window. No second chance. The governance must be pre-inference — or it doesn't exist.
Verify the hardware producing inference is real — not spoofed, not compromised
Certify before the output executes — there is no after in zero gravity
Receipted chain of custody for every autonomous decision
Already tested. Satellite relay governance. Autonomous de-orbit inference. Edge AI under the 1 Joule Rule. The simulations ran. The receipts exist.