Nexus Synergy Engagement Package — E09 · France — DGA + Marine nationale 'Maîtrise des Fonds Marins' (MFM) seabed-warfare ecosystem; defensive CUI-protection slice only; Observatoire MFM (FMES Toulon) + GISMER/CEPHISMER; prime-gatekept (Thales/Naval Group/Exail)

STANDING BANNER — read first. (1) Status honesty: Nexus Synergy Ltd is pre-incorporation, pre-revenue, pre-pilot. France's DGA, the Marine nationale, the Observatoire Maîtrise des Fonds Marins, GISMER/CEPHISMER and the SEACURE/France-2030 primes are TARGETS, not customers, partners, or pilots. The ~178-surface UI gallery is a gallery, not traction. This package is explicitly DISCOVERY-ONLY / LOW-CONFIDENCE: the buyer is real and well-documented, but there is no named warm route into a sovereignty-protected, prime-gatekept French national-defence programme — the score below carries a wide σ and is a prioritisation prior, not a forecast. (2) Anti-cookie-cutter: the §8 mockups are generated programmatically by _build/archetypes/gen_france-seabed-infra-protection.py and filled with this buyer's real entities (Golfe de Gascogne / Toulon-approach CUI corridor, EMODnet −2400 m, SHOM ENC, Exail-A6K-M AUV, TravOcean ROV, GISMER/CEPHISMER, the MFM "know/watch/act" functions) — these are this buyer's screens. (3) All probabilities are [PRIOR] — every factor cell carries a dated rationale and re-scores after the first real conversation.

---

§0 — Header + one-line thesis

France Maîtrise des Fonds Marins — a sovereign-FR, EU-only decision/COP layer for the defensive "know + watch" critical-undersea-infrastructure (CUI) slice of the French seabed-warfare programme.

Thesis. France has the most mature national seabed programme in Europe: a dated ministerial Maîtrise des Fonds Marins strategy (14 February 2022), a dedicated Observatory (April 2026), France-2030 deep-sea robotics (Exail AUV + TravOcean ROV to 6 000 m) and a pôle d'excellence grands fonds marins (Dec 2025). What that programme builds is sensors, AUVs, ROVs and platforms; what it consistently lacks is a vendor-neutral, sovereign decision / common-operating-picture layer that fuses AIS + EMODnet bathymetry + SHOM ENC + DAS acoustic + magnetometer + AUV/ROV telemetry into one human-gated, fully-audited CUI-protection picture — with raw capture staying inside France and only derived products + survey-cues crossing into operational hands. Nexus Synergy is that layer, framed strictly to the defensive "know + watch" functions of the strategy — never its explicit "act"/offensive (mining, special-action) component.

---

§1 — Entity snapshot (cited)

Important framing. "France DGA" is not one buyer that signs a SaaS cheque. Realistic entry, if any, is (a) as a sub-supplier / module inside a French-prime-led programme (SEACURE-adjacent, Naval-Group/Thales/Exail), and/or (b) via the Agence de l'Innovation de Défense (AID) innovation route (Guichet Unique) — both of which strongly favour French and French-led-EU industry; eligibility of a non-French, pre-incorporation firm is [unverified].

---

§2 — The pain (web-verified, dated, cited)

1. A live, repeated forcing function on European CUI. The Baltic Sea saw a cluster of subsea incidents — the Estlink 2 Finland–Estonia power cable severed 25 Dec 2024 (anchor-drag, "shadow-fleet" tanker), and Nov-2024 fibre cuts (Lithuania–Sweden, Finland–Germany) — and NATO Baltic Sentry launched Jan 2025 in response. France had already unveiled its seabed-warfare strategy on 14 Feb 2022 (months before Nord Stream); the Nord Stream attacks (Sep 2022) then sharply elevated attention and prompted France to dive deeper — accelerate and broaden — its undersea-security effort. [verified — en.wikipedia.org/wiki/2024_Estlink_2_incident, usnews.com 2022/10]
2. France made it a sovereign capability programme. The Maîtrise des fonds marins strategy (14 Feb 2022) commits France to a sovereign ability to know, watch and act on the seabed to 6 000 m, explicitly to protect CUI; the France-2030 "invest in the seabed" objective and the DGA/SGPI 27-Sep-2024 AUV/ROV order operationalise it; a bilan trois ans (Feb 2025) and the Observatoire (Apr 2026) show sustained momentum. [verified — defense.gouv.fr, opex360.com 2025/02]
3. The capability gap is fusion + sovereign decision, not more robots. France is procuring platforms and sensors (Exail AUV, TravOcean ROV, GISMER robots, SEACURE system-of-systems by 2028). What is consistently missing is a vendor-neutral, conformity-by-construction decision/COP layer that fuses heterogeneous outputs into one human-gated CUI picture without raw national capture leaving French sovereignty. NATO's JEF Nordic Warden (AI vessel-risk, Jan 2025) shows the demand for exactly this fusion layer at coalition scale. [verified — atlanticcouncil.org]
4. A decline-policed adjacency the buyer itself carries. France's strategy was confirmed in Mar 2025 to include an offensive ("volet offensif") component — future robots, naval special actions, mining. That is precisely the boundary Nexus Synergy must not cross: our slice is the defensive "know + watch" CUI-protection decision layer only. [verified — navalnews.com 2025/03]

The pain in one line. France has the platforms, the budget, the mandate and the Observatory — but no sovereign, conformity-by-construction decision/COP layer that fuses them into one human-gated CUI-protection picture while keeping raw capture inside France.

---

§3 — Use-case & value (DECLINE-SAFE framing)

The use case: a sovereign-FR seabed CUI-protection COP for the Golfe de Gascogne / Mediterranean-approach cable-and-interconnector corridors, fusing AIS + EMODnet bathymetry + IHO/SHOM S-101 ENC + DAS acoustic (fibre-tap) + magnetometer (anchor-drag) + AUV/ROV survey telemetry into one human-gated picture, with a 6-persona Decision Room seating a civil-society reviewer and a warrant authority inside the decision, raw staying sovereign-FR (only derived products + survey-cues leave the tenancy), and sealed attribution revealable only under dual-control + a warrant instrument.

The value: earlier, defensible detection of a vessel-proximity threat to a French power interconnector or fibre trunk; a single auditable picture that cues an AUV/ROV survey pass (the "watch" function) rather than a manual scramble; conformity-by-construction (FRIA + provenance ledger + transparency portal) for the EU AI Act / NIS2 framing around CUI tooling.

DECLINE-SAFE boundary (explicit and load-bearing — this buyer is decline-sensitive): this is the defensive civil-security "know + watch" CUI-protection slice ONLY. It is NOT the French strategy's "act"/offensive component (no mining, no kinetic, no special-action tasking); it is NOT warfighting; it is NOT predictive policing, biometric identification, emotion recognition, social scoring, untargeted scraping, or psychometric profiling. The picture is object/vessel-track-centric, human-gated, fully audited; raw stays sovereign-FR; flag-state attribution stays sealed pending a warrant; decisions are survey-tasking cues (cue an AUV/ROV to look), never live targeting.

---

§4 — Ontology (this buyer's domain entities + relationships)

CuiSegment ──route_segment→ BathymetryTile           (EMODnet / GEBCO / SHOM S-101 ENC)
CuiSegment ──asset_kind→ {power_interconnector, fibre_trunk, gas_pipeline}
CuiSegment ──monitored_by→ SensorAsset {DAS, magnetometer, hydrophone}
SensorAsset ──deploys→ SensorSignature {das_acoustic, em_magnetometer}
SensorSignature ──coincident_with→ VesselOfInterest    (AIS-dark / RF-dark loiter)
VesselOfInterest ──adjacency(km, min)→ CuiSegment      (0.9 km / 31-min window)
SurveyAsset {AUV(Exail-A6K-M), ROV(TravOcean), GISMER-robot} ──cued_to→ CuiSegment  (watch only)
WatchFunction {KNOW, WATCH} ──authorises→ SurveyAsset   (NEVER the ACT/offensive function)
DerivedProduct ──derived_from→ SensorSignature ──released_to→ CMF_NOC  (advisory)
Attribution ──sealed_by→ WarrantInstrument + DualControl

Entity resolution across the French sensor estate uses Fellegi–Sunter match weights (time-correlation, positional cross-track, spectral signature) so a DAS-array observation and an AUV-survey hit resolve to one event — mirroring the W-11 / W-15 sub-sea worked examples (W ≈ 18.5 in the high-stakes Tier-4 case, T_μ = 6.0). PROV-O records which French sensor produced which observation, and a hard ontology constraint binds SurveyAsset only to the KNOW/WATCH functions — the ACT/offensive function is not a node Nexus Synergy can task.

---

§5 — Data model (synergy.* tables, RLS, load-bearing CHECK)

France is its own tenant (org_id = app.current_org_id). The load-bearing safety properties: (1) no asset may ever be cued to the offensive "act" function from this layer — only the defensive KNOW/WATCH survey functions; (2) no flag-state attribution reveal without BOTH dual-control AND a warrant. Both are enforced as DB CHECKs, not application logic.

SQL
43 lines
Copy
-- Per-segment CUI asset, scoped to the owning sovereign tenant (FR).
CREATE TABLE synergy.cui_segment (
    id               uuid PRIMARY KEY DEFAULT gen_random_uuid(),
    org_id           uuid NOT NULL,                         -- RLS: app.current_org_id (sovereign FR)
    segment_uid      text NOT NULL,                         -- e.g. corridor KP-188
    operator         text NOT NULL,                         -- NIS2 / OIV essential entity
    asset_kind       text NOT NULL CHECK (asset_kind IN ('power_interconnector','fibre_trunk','gas_pipeline')),
    corridor         geography(LINESTRING,4326) NOT NULL,   -- route on the globe
    depth_m          int,                                   -- e.g. -2400
    enc_ref          text NOT NULL,                         -- SHOM / IHO S-101 ENC reference
    unclos_regime    text NOT NULL DEFAULT 'continental_shelf',  -- UNCLOS Art 113
    UNIQUE (org_id, segment_uid)
);
ALTER TABLE synergy.cui_segment ENABLE ROW LEVEL SECURITY;
CREATE POLICY org_isolation ON synergy.cui_segment
    USING (org_id = current_setting('app.current_org_id')::uuid);

-- A survey-tasking cue. The decline-safe doctrine (defensive WATCH only) lives here.
CREATE TABLE synergy.survey_cue (
    id                 uuid PRIMARY KEY DEFAULT gen_random_uuid(),
    org_id             uuid NOT NULL,
    cui_segment_id     uuid NOT NULL REFERENCES synergy.cui_segment(id),
    seabed_function    text NOT NULL CHECK (seabed_function IN ('know','watch')),  -- NEVER 'act'
    survey_asset       text NOT NULL CHECK (survey_asset IN ('auv_exail_a6km','rov_travocean','gismer_robot')),
    anomaly_score      numeric(4,3),                        -- e.g. 0.83
    proportionality    numeric(4,3),                        -- N*P*L envelope (~0.83)
    dual_signoff_id    text,                                -- 2/2 dual-control; NULL until signed
    warrant_instrument_id text,                             -- NULL unless attribution reveal sought
    released_at        timestamptz,                         -- NULL until human-gated release (advisory)
    flag_state_revealed_at timestamptz,                     -- NULL unless dual-control + warrant
    retain_until       date NOT NULL,                       -- retention per FR data-protection rule
    prov_o             jsonb NOT NULL,
    -- LOAD-BEARING #1 (decline-safe): a cue may only be released for a DEFENSIVE survey function,
    -- and only after a human dual-control sign-off. The 'act'/offensive function is structurally absent.
    CHECK (released_at IS NULL
           OR (seabed_function IN ('know','watch') AND dual_signoff_id IS NOT NULL)),
    -- LOAD-BEARING #2: no flag-state attribution reveal without BOTH dual-control AND a warrant.
    CHECK (flag_state_revealed_at IS NULL
           OR (dual_signoff_id IS NOT NULL AND warrant_instrument_id IS NOT NULL))
);
ALTER TABLE synergy.survey_cue ENABLE ROW LEVEL SECURITY;
CREATE POLICY org_isolation ON synergy.survey_cue
    USING (org_id = current_setting('app.current_org_id')::uuid);

CHECK #1 makes the decline-safety physically enforced at the database: a cue can only ever be a know/watch survey and only after a human gate — the offensive "act" function is not even a permitted value, so the layer cannot be repurposed to task a kinetic/mining action by a UI or application bug. CHECK #2 blocks any attribution reveal without dual-control + warrant.

---

§6 — Action-loop pseudocode (detect → enrich → triage → gate → task → execute → BDA → close)

Python
42 lines
Copy
def run_cui_watch(segment_uid, requesting_org):
    # 1. DETECT — AIS-dark / RF-dark loiter astride a French CUI corridor.
    seg = cui_segment(segment_uid)                               # corridor KP-188, -2400m
    voi = detect_loiter(seg, radius_km=0.9, window_min=31)       # VOI RF-dark
    if voi is None:
        return Disposition.MONITOR

    # 2. ENRICH — fuse DAS acoustic tap + magnetometer EM anchor-drag (derived only).
    das = das_signature(seg, window_min=31)                     # tap-pattern @ KP-188
    em  = em_signature(seg)                                      # anchor-drag
    ev  = fuse_event(seg, inputs="derived_only")                 # raw stays sovereign-FR
    if ev is None or ev.fs_link_weight < T_MU_TIER4:            # W < 6.0 -> no event
        return Disposition.MONITOR

    # 3. TRIAGE — classify; reject benign explanations before escalation.
    cls = classify(das, em, voi)                                # anchor_drag .83 vs trawl_snag .11
    if cls.benign_prob > 0.5:
        return Disposition.MONITOR                               # lawful fishing / transit

    # 4. GATE — Art-8-ECHR proportionality; DEFENSIVE WATCH only; HUMAN sign-off required.
    P = proportionality(necessity=0.92, proportion=0.87, least_intrusive=0.80,
                        scope="defensive_watch")                # P = 0.83
    decision = decision_room(personas=[
        "op_tasking","proportionality","civil_society",
        "devils_advocate","subsea_expert","warrant_authority"]) # 6 personas
    if not decision.dual_signed():                              # 2/2 dual-control
        return Disposition.HOLD                                  # no release, no cue

    # 5. TASK — DEFENSIVE survey cue ONLY (know/watch); attribution stays SEALED.
    cue = survey_cue(function="watch", asset="auv_exail_a6km",  # NEVER 'act'/offensive
                     note="cue AUV survey pass; no live op, no targeting")
    seal_attribution(ev)                                        # reveal-under-warrant only

    # 6. EXECUTE — advisory write-back through nexus-workflows to the CMF NOC.
    job = workflows.submit("cue_auv_survey", payload=cue, dual_control=True)

    # 7. BDA — did the survey cue validate the fusion + the AUV-handoff sync-ladder?
    bda = assess(job, criteria=["fusion_validated","auv_handoff_sync_ok"])

    # 8. CLOSE — record PROV-O chain; lodge derived product (FR retention rule).
    close_case(seg, ev, decision, bda, retain="fr_dp_rule")
    return Disposition.CLOSED

---

§7 — nexus-workflows YAML DAG (same loop, declarative, blocking human-gate)

YAML
36 lines
Copy
# nexus-workflows job: cui_watch_loop  (defensive 'know'+'watch' CUI protection only)
name: cui_watch_loop
trigger: { on: das_or_em_anomaly, scope: "org=app.current_org_id" }
nodes:
  - id: detect_loiter
    run: nexus_synergy.subsea.detect_loiter        # AIS/RF-dark, 0.9km/31min
  - id: enrich_fuse
    needs: [detect_loiter]
    run: nexus_synergy.subsea.fuse_event            # DERIVED ONLY; raw stays sovereign-FR
    params: { inputs: derived_only, fs_threshold: 6.0 }
  - id: triage_classify
    needs: [enrich_fuse]
    run: nexus_synergy.subsea.classify              # anchor_drag vs benign
  - id: HUMAN_GATE                                  # <-- BLOCKING node
    needs: [triage_classify]
    type: human_decision
    blocking: true
    quorum: 2                                       # 2/2 dual-control sign-off
    personas: [op_tasking, proportionality, civil_society,
               devils_advocate, subsea_expert, warrant_authority]
    allowed_functions: [know, watch]                # offensive 'act' function NOT permitted
    refuse_path: hold_and_log                       # dissent -> HOLD, no release/no cue
  - id: task_survey_cue
    needs: [HUMAN_GATE]
    run: nexus_synergy.subsea.survey_cue            # DEFENSIVE survey cue only; attribution SEALED
    params: { function: watch, asset: auv_exail_a6km }
  - id: execute_writeback
    needs: [task_survey_cue]
    run: nexus_synergy.workflows.callback_post      # advisory write-back to CMF NOC, dual-control
    params: { targets: [cmf_noc_notify, nis2_register] }
  - id: bda_assess
    needs: [execute_writeback]
    run: nexus_synergy.subsea.assess
  - id: close_provo
    needs: [bda_assess]
    run: nexus_synergy.audit.close                  # PROV-O chain, FR retention rule

---

§8 — UI/UX mockups (verbatim generated ASCII)

Figure 8.1 — globe_rmp (PRIMARY). ShellLayout + TopBar (with RendererChip = CesiumJS, PccPill = 4-sync) + ClassificationBanner (EU-RESTRICTED / TLP:AMBER) + LeftSidebar glyph rail + MapConsole rendering the CesiumJS under-water-tilt globe over the French CUI corridor + a right-rail CONTACTS list with the ChatTerminal Decision-Room transcript + BottomStatusBar. Generated by gen_france-seabed-infra-protection.py.

Figure — Recognised maritime picture (globe_rmp). Production-fidelity React surface (buildable); the faithful ASCII follows.

+--------------------------------------------------------------------------------------------------+
| French CUI Seabed RMP - Maitrise des Fonds Marins defensive watch (CesiumJS-globe)               |
+-------------------------------------------------------------------+------------------------------+
|         . - '' - .       GOLFE DE GASCOGNE / MED approaches       | CONTACTS (sorted: sub-sea)   |
|      /   o ----._  \   CUI: power interconnector + fibre trunk    | VOI vessel  RF-DARK          |
|     |  [!] AIS dark |   EMODnet bathy -2400m :: SHOM ENC base     |   AIS gaps, loiter 31m       |
|     |   VOI contact  |  VOI loiter (AIS gaps, RF-dark)            |   0.9km of corridor KP-188   |
|     |  loiter 31 min |   astride cable corridor SW of Toulon      | DAS acoustic event           |
|      \ __X DAS tap __/   DAS acoustic tap-pattern @ KP-188        |   KP-188  tap-pattern        |
|       \ '. _____ .'/    magnetometer EM anchor-drag (consistent)  |   class: anchor_drag .83     |
|         ' - x__ - '     AUV Exail-A6K-M / ROV TravOcean @6000m    | EM magnetometer              |
|   Camera {lat 43.0 lon 5.9 alt 600km} CesiumJS-globe  L4          |   anchor-drag consistent     |
|   Concurrence: AIS-dark + DAS tap + EM drag, 0.9km/31min          | [#] MFM defensive watch      |
|   -> ANOMALY 0.83  ::  HUMAN GATE :: attribution SEALED           |   GISMER/CEPHISMER Toulon    |
|                                                                   | Function: KNOW + WATCH       |
|                                                                   |   NOT 'act'/offensive        |
|                                                                   | gate: CMF alert + AUV cue    |
|                                                                   |   survey-task (no live op)   |
|                                                                   | Attribution: SEALED          |
|                                                                   |   reveal-under-warrant       |
+-------------------------------------------------------------------+------------------------------+
| EMODnet bathy + SHOM ENC + DAS + magnetometer + AIS :: MFM defensive CUI watch, survey-tasking   |
+--------------------------------------------------------------------------------------------------+

Figure 8.2 — command_dashboard (SECONDARY). The MFM-programme readiness board: ShellLayout + KPI tiles (MFM strategy / France-2030 robots / pôle d'excellence) + a LIVE ALERT FEED carrying the dated mandate timeline and the decline-safe doctrine. Generated by the same script.

Figure — Command dashboard (KPI + live feed). Production-fidelity React surface (buildable); the faithful ASCII follows.

+--------------------------------------------------------------------------------------------------+
| Maitrise des Fonds Marins Programme Readiness - DGA/Marine nationale (command_dashboard)         |
+--------------------------------------------------------------------------------------------------+
| MFM strategy (14 Feb 2022)     | France 2030 deep-sea robots    | Pole d'excellence GFM          |
| know / watch / act to 6000m    | DGA+SGPI order 27 Sep 2024     | launched FID Dec 2025          |
| Observatoire MFM Apr 2026      | Exail AUV + TravOcean ROV      | prog ~EUR300m (~100m milestone)|
+--------------------------------------------------------------------------------------------------+
| LIVE ALERT FEED                                                                                  |
|  14 Feb 2022  Ministerial Maitrise des Fonds Marins strategy adopted (DGA)                       |
|  27 Sep 2024  DGA+SGPI order AUV(Exail A6K-M)+ROV(TravOcean) to 6000m                            |
|  Dec 2025     Pole d'excellence grands fonds marins launched at FID 2025                         |
|  24 Apr 2026  Observatoire MFM operational (FMES Toulon, RAdm Mallard)                           |
|  scope      DEFENSIVE 'know'+'watch' CUI slice; NOT the 'act'/offensive                          |
|  decision   6-persona Decision Room incl. civil-society + warrant auth                           |
|  sharing    raw stays sovereign-FR; derived/advisory products only                              |
|  gate       survey-tasking cue to AUV/ROV; NO live op; SEALED attr                              |
|  decline    civil-security CUI protection; NOT warfighting; no biometric                         |
+--------------------------------------------------------------------------------------------------+

(Bespoke SVG of the primary screen: _build/figures/france-seabed-infra-protection/uc-globe_rmp.svg — same globe_rmp archetype, real component names.)

---

§8b — Field-unit (Pixel) surfaces

The same scenario on the Pixel 10 Pro Fold field unit (Nexus Field app), tightly coordinated with the dashboard COP above — command pushes the task, the unit accepts + ACKs, shares position and reports back to the COP. Built on the same synergy.field_unit / field_task / field_report contract; see §9 and the cluster coordination composite.

Figure §8b.1 — Folded cover · tasking glance (ground_glance): the incoming IMMEDIATE task, ACCEPT + ACK, alert chips, bearing-to-objective.

Figure §8b.2 — Unfolded inner display · field COP: two-pane mini-map + task list + teammate roster + air/command coordination + PTT, with the Material-3 NavigationBar + Report FAB.

§9 — UI/UX flow (literal click-path)

Click-path: Analyst opens MapConsole (CesiumJS, under-water tilt) over the Golfe-de-Gascogne / Toulon-approach corridor → an AIS-dark VOI contact raises a red pulse astride the CUI corridor (KP-188) → click the VOI in the CONTACTS rail → the DAS tap-pattern (KP-188) and EM anchor-drag correlate into one event (Fellegi–Sunter) → the anomaly score (0.83) and proportionality envelope (N .92 · P .87 · L .80) populate → the Decision Room opens inside ChatTerminal with 6 personas → the analyst reads the civil-society dissent (derived-only release, raw stays sovereign-FR) and the warrant authority verdict (attribution SEALED) → /dual-control sign (2/2) unlocks /execute → AUV survey cue (defensive WATCH only) → the write-back fires a nexus-workflows job to the CMF NOC; the BottomStatusBar shows the EU-sovereign-only AI Provider Router + raw-stays-sovereign-FR state.

MapConsole(VOI red pulse) ──▶ CONTACTS(select VOI) ──▶ fuse DAS+EM (Fellegi-Sunter)
        │                                                      │
        ▼                                                      ▼
  anomaly 0.83 + N/P/L envelope ──▶ ChatTerminal Decision Room (6 personas)
                                            │
        civil-society ──▶ derived-only │ warrant auth ──▶ SEALED
                                            ▼
            /dual-control sign (2/2) ──▶ /execute AUV survey cue (WATCH only) ──▶ nexus-workflows
                                            ▼
                       write-back (CMF NOC + NIS2 register) ──▶ PROV-O close

---

§10 — Decision-Room transcript (the gated decision)

Scenario: A defensive CUI-watch event in the Golfe-de-Gascogne / Toulon-approach corridor. A DAS tap-pattern on a power interconnector / fibre trunk (KP-188, −2400 m) + a magnetometer anchor-drag + an RF-dark vessel loitering 0.9 km / 31 min. The gate question: cue an AUV survey pass (the defensive "watch" function), how much may be released to the CMF NOC, and whether attribution may be revealed.

▣ EU-RESTRICTED · TLP:AMBER · MFM DEFENSIVE WATCH (survey-tasking only) ▣

OP-TASKING [L4]            DAS tap + EM anchor-drag + RF-dark adjacency,
                          0.9 km / 31 min astride CUI corridor KP-188.
                          Recommend: cue AUV Exail-A6K-M survey pass;
                          alert the CMF NOC. (WATCH function, no live op.)

PROPORTIONALITY [L4]      Envelope-strict: N 0.92 · P 0.87 · L 0.80 -> P = 0.83.
                          Defensive 'watch' scope; no targeting, no kinetic.
                          Art-8-ECHR proportionality satisfied for a survey cue.

CIVIL-SOCIETY [L4]        DISSENT on default sharing AND on scope-creep.
                          (1) Release DERIVED-attribution only to the CMF NOC;
                          raw IQ stays sovereign-FR. (2) This layer must NEVER
                          touch the strategy's offensive 'act' function -- confirm
                          the 'watch'-only enum is enforced.   <-- changes the rec

DEVIL'S ADVOCATE [L4]     Counter: anchor-drag could be a lawful fishing vessel
                          dragging gear over the corridor. Do NOT reveal flag-state
                          on acoustic + EM alone; do NOT escalate beyond survey.

SUBSEA EXPERT [L4]        Tap-pattern + anchor-drag over a power interconnector at
                          -2400 m is not benign loiter; consistent with the post-
                          2022 CUI-threat profile (Nord Stream / Estlink-2).
                          ObservationLevel = Direct.

WARRANT AUTHORITY [L4]    VERDICT: confirm 'watch'-only enum (CHECK #1) -- the
                          offensive 'act' function is structurally absent from this
                          layer. Flag-state attribution stays SEALED; reveal needs
                          2/2 dual-control AND a warrant instrument (CHECK #2).
                          GATE: ACCEPT a DEFENSIVE AUV survey cue at P=0.83.
                          Attribution: AWAIT.

[ /dual-control sign -- analyst ]   [ /dual-control sign -- CMF watch-officer ]
[ /execute -> AUV survey cue (WATCH only) ]   [ Refuse + log ]

Decision-changing dissent: the civil-society persona's two-part intervention (a) narrowed the default from raw-IQ release to derived-attribution-only to the CMF NOC, and (b) forced an explicit re-confirmation that the layer's watch-only enum holds — so the survey cue can never silently widen into the strategy's offensive "act" function. The recommendation was then dual-signed at P = 0.83 with attribution sealed.

---

§11 — Write-back + BDA + PROV-O chain + deltas-only regulatory traceback

Write-back. On /execute, a nexus-workflows job (cue_auv_survey) posts an advisory survey cue to (a) the CMF NOC (watch-officer cues the Exail-A6K-M / TravOcean asset) and (b) the NIS2 register (operator notified of a CUI-segment anomaly). It is a survey cue — a request to look — never a live or kinetic tasking.

BDA (loop close). The cue validates two things the discovery/exercise phase exists to prove: the DAS+EM Fellegi–Sunter fusion resolved heterogeneous French-sensor events into one event, and the AUV-handoff sync-ladder held under emulated DEGRADED-comms (EMCON / Galileo-PRS). Failures (e.g. an AUV telemetry feed not yet wired) are surfaced as explicit Phase-2 dependencies, not hidden.

PROV-O chain. wasDerivedFrom: the survey cue ← (DAS signature ⊕ EM signature ⊕ VOI track); wasAttributedTo: each observation → the specific French sensor that generated it; wasAssociatedWith: the survey cue → the 6-persona Decision Room + the 2/2 dual-control; wasGeneratedBy: the cui_watch_loop workflow run. The chain is merkle-anchored.

Deltas-only regulatory traceback (only what differs from the §4.0 defaults): AI Act — high-risk critical-infrastructure context ⇒ FRIA on file; no Article-5 prohibited practice (no biometric, no social scoring); EU AI Act Article 5 applicable 2 February 2025. GDPR/FR data-protection — derived-only release, EU-hosted, sovereign-FR raw retention. NIS2 — the CUI operator is an essential entity (FR transposes via the "résilience des activités d'importance vitale" regime); Art-23 incident register updated. UNCLOS Art 113 — continental-shelf regime recorded per segment. Defensive-scope invariant — DB CHECK #1 records that no cue ever left the know/watch functions.

---

§12 — Buyer & stakeholders

Honest note: the only approachable named entity here is the Observatoire / FMES (an open think-tank), and that is a research/discovery conversation, not a procurement path. There is no warm route to DGA/Marine nationale in the corpus.

---

§13 — Competition / incumbency + comparator (cited)

White space (contested): a sovereign-FR, EU-only, conformity-by-construction decision/COP layer above the French primes' sensors. Genuine seam — but France's strong sovereign-industrial preference and prime gatekeeping make it narrow and hard-to-reach for a non-French firm. This is why the package is discovery-only.

---

§14 — Readiness (honest, pre-pilot)

Reuse (real): the globe_rmp seabed surface, the DAS/EM correlation pattern, the 6-persona Decision Room (incl. civil-society + warrant authority), the watch-only CHECK invariant, and the cui_watch_loop workflow are all already specified in the UC-55 / W-11 / W-15 worked examples and the archetype library. The AI Provider Router (EU-sovereign model routing) and the FRIA / provenance-ledger / transparency-portal conformity components exist as architecture.

Real gaps (honest): (1) no named route into DGA / Marine nationale — the entire commercial premise is unproven; (2) no live French sensor data — DAS/magnetometer/AUV feeds are notional; (3) French sovereign-industrial preference + prime gatekeeping means a non-French firm likely enters only as a sub-supplier, if at all; (4) AID / Guichet-Unique eligibility for a non-French, pre-incorp firm is [unverified]; (5) language/locale (French-domain UI, SHOM ENC integration) is unscoped; (6) L4 / "Diffusion Restreinte" accreditation is unbuilt. This is a long-horizon DISCOVERY / WATCH-LIST opportunity, not a near-term deployment.

---

§15 — ENGAGEMENT PLAYBOOK

Reality check first: because there is no warm route, Stage 1 here is intelligence-gathering and an EU-consortium back-door, not a direct DGA approach. The honest first move is to earn a delivered Irish/Belgian decline-safe reference (NDFEM/MAHI) and reach France via an EDF/EUDIS consortium where a French prime needs a sovereign COP module — not to cold-pitch DGA.

60-min discovery agenda (adapted): 0–5 frame + the decline-safe statement (defensive "know/watch" CUI only; never the offensive "act" function); 5–20 "walk me through a CUI-anomaly from detect to close — how many systems, who decides, how do you cue an AUV/ROV?"; 20–35 lawful basis + AI-Act/NIS2 scope + where does the data and model run (sovereign-FR); 35–45 the route (FMES research vs French-prime sub-supplier vs EDF consortium); 45–55 map their loop onto detect→enrich→triage→gate→task(survey-cue)→execute→assess→close; 55–60 reflect + agree the demo date. Bring no slides.

1-page proposal (filled): Title — "Sovereign-FR defensive CUI seabed-watch decision layer — a module for the MFM/SEACURE ecosystem." Sponsor — a French prime (SEACURE-adjacent) or FMES research host, with MAHI as the EU contracting vehicle. The one thing we prove — "we fuse DAS + AIS + EM into one CUI event and reach a human-gated, derived-only, sealed-attribution survey cue to an Exail-A6K-M AUV, with raw never leaving French sovereignty and the offensive 'act' function structurally absent." Scope IN — globe_rmp + Decision Room + watch-only CHECK + workflow write-back. OUT — the offensive "act" function, live/kinetic tasking, classified strands, any biometric/predictive (Declined-List + offensive-scope excluded). Data — representative/exercise DAS/AIS/EM, EU-hosted. Success criteria — fusion validated; AUV-handoff sync-ladder holds under DEGRADED-comms emulation; 6-persona gate audit complete (baselines week 1). Timeline — 8–12 wks. Commercials — €0 to sponsor if grant-funded (EDF/EUDIS); production band Tier-1 €120–240k/yr to Tier-3 €2.4–4.8m/yr only if asked. After — MOU + EDF-2026 consortium intent.

LOI checklist (key items): parties (prime/MAHI vehicle pre-incorp / Founder in own name under CA 2014 s.45); non-binding statement; scope from proposal; success criteria referenced; named sponsor; indicative timeline; "to be scoped under EDF-2026 / a French-prime sub-contract"; conditions (funding, accreditation, data-sharing); binding vs non-binding stated; mutual NDA; Irish governing law; Declined-List + defensive-watch-only affirmation; signatures; ⚖️ solicitor review confirmed.

Warm-intro email (adapted draft #3, reframed for the research/consortium door): Subject: "A sovereign-EU defensive seabed-CUI decision layer — a research conversation." Body: France leads Europe on seabed mastery (MFM strategy, the Observatory, France-2030 robots); the consistent gap is a vendor-neutral, conformity-by-construction decision/COP layer that fuses the sensors into one human-gated CUI-watch picture, raw staying sovereign-FR. We publicly refuse the offensive "act" function and every Declined-List capability. Ask: a 30-minute research conversation (FMES/Observatoire) and an introduction to where a French-prime EDF consortium might use such a module.

---

§16 — PM / timeline

                 2026                          2027                  2028
        Q3      Q4      Q1      Q2      Q3      Q4      Q1      Q2
PRE: deliver IE/BE reference  [#####]    (gates any French credibility)
S1 indirect contact (FMES/prime)    [##]
S2 free discovery (research)           [###]
S3 demo                                   [##]
S4 EDF/prime module pilot                   [######]
S5 LOI/MOU                                       [##]                     <- Patrick-check
EDF-2026 bid window  [====29 Sep 2026 close====]                          <- external clock
S6 consortium/sub-contract                          [###################]
   (gated by EDF award + French-prime sponsor + Stamp-4/partner vehicle)

Milestones: M0 delivered IE/BE decline-safe reference (Q4-26, the real precondition); M1 FMES/prime indirect contact (Q1-27); M2 discovery PASS (Q1-27); M3 demo accept (Q2-27); M4 module exercise report (Q3-27); M5 MOU (Q4-27); M6 EDF/prime award (2028+). Critical path: two gates compound — (a) the reference gate (no French sovereign buyer will look at a non-French pre-incorp firm with zero delivered reference), and (b) the Stamp-1G / partner-vehicle gate (a paid sub-contract cannot be signed by the solo founder; a French prime or MAHI as contracting entity is required, lifting V 0.10→0.75), alongside the EDF-2026 29 Sep 2026 external clock.

RACI:

---

§17 — Funding / procurement vehicle

• EDF-2026 — €1bn WP2026; seabed topic family (e.g. …-UWW-CSBI "Layered critical seabed infrastructure protection"); calls close 29 September 2026; consortia ≥3 entities in ≥3 Member States (non-dilutive) — the most realistic back-door to France via a French-prime-led consortium. [verified]
• EUDIS — SME calls up to €4M research / €6M development. [verified]
• Agence de l'Innovation de Défense (AID) — Guichet Unique innovation desk; non-French / pre-incorp eligibility [unverified] — must be confirmed before relying on it. [verified channel]
• France 2030 "invest in the seabed" — DGA/SGPI vehicle that funded the Exail/TravOcean robots; a national instrument, almost certainly French-industry-bound — not directly accessible. [verified]
• Blocked: NATO DIANA — Ireland is non-NATO; ineligible. Use EU vehicles only. [verified — diana.nato.int]

---

§18 — TWO-STAGE FORMULA SCORECARD

Confidence flag: this row is DISCOVERY-ONLY / LOW-CONFIDENCE (σ ≈ 0.50, cold). A named buyer and a procurement route are established from real sources, so the row is scored (not null), but the dominant driver — no warm route into a sovereignty-protected, prime-gatekept French buyer — keeps Stage-1 access_warmth, white_space, and cycle_speed low. Re-score the instant any real French-side conversation happens.

Stage-1 {mandate_pull, access_warmth, demonstrability, decline_safety, white_space, cycle_speed, pillar_fit} — all [PRIOR]:

S1 = 3.18 → P_LOI = 0.55/(1+exp(−1.15·(3.18−2.6))) = 0.364 (≈ 36% [PRIOR]). Score100 = S1·20 = 63.6.

Stage-2 {contractability, funding_to_pay, procurement_clarity, incumbency_displacement, time_to_value, reference_leverage} — all [PRIOR]:

S2 = 1.68 → P_raw = 0.70/(1+exp(−1.1·(1.68−2.8))) = 0.158 [PRIOR].

Legal gate G = V·T, with t_resolve = 18 mo, H_conv = 30, so T = 0.400:

Reconciliation: board_anchor is [PRIOR-derive], value high. Score100 = 63.6 sits in the medium-high band — above the weakest EU rows (HSE 56.1, NCSC 58.1) and consistent with the slower, prime-gatekept maritime rows (ComReg 62.2, Coast-Guard 60.6, Dublin-Port/EMSA ~60.6), but below the strongest-mandate rows because France's sovereignty barrier crushes Stage-2. The mandate is the strongest in the volume; the route is the weakest — which is exactly why this is DISCOVERY-ONLY. Even at V=1.0, P(Contract) is only ~2.3%, dominated by the Stage-2 access/contract collapse — the honest read is watch-list / EDF-consortium back-door, not a direct pursuit.

---

§19 — Commercial

Requirements: EU-hosted sovereign infra; French "Diffusion Restreinte" / L4 accreditation for any classified strand; a French prime or MAHI as the contracting entity; DPA + data-sharing agreement; FRIA on file; the watch-only enum affirmed in contract. Cost advantage from open-source substrate + sovereign HPC + shared conformity file + no forward-deployed-engineer dependency — but this is a price only reachable via a prime/consortium, not a direct French SaaS sale.

---

§20 — Legal blockers

---

§21 — Warm-intro contact + the specific ask

There is no warm intro to the buyer. The honest asks are indirect:

• Pieter-Jan Note (MAHI CEO) — warm, founder-to-founder. Ask: whether a future EDF-2026 seabed consortium could include a French prime that needs a sovereign defensive CUI-watch decision module, with MAHI as the EU contracting vehicle — and whether MAHI has any French-prime relationships to bridge. [verified person; email not public]
• Observatoire MFM / FMES (RAdm (ret.) Jacques Mallard) — open think-tank, not a buyer. Ask: a research conversation on the sovereign CUI-decision-layer gap (strictly the defensive "know/watch" slice), to validate the thesis and surface where a French prime might use such a module. [verified person; FMES is publicly approachable]
• Patrick Walsh (Dogpatch) — warm. Ask: not a France intro (out of his network) but a reminder that the delivered Irish reference (NDFEM/MAHI) is the precondition that makes any France conversation credible. [verified — contact-register]

---

§22 — Open questions + consolidated Sources

Open questions: (1) Is there any path for a non-French firm to contribute a module to the MFM/France-2030 programme other than as a French-prime sub-supplier? [TBD] (2) Is AID / Guichet-Unique open to a non-French, pre-incorporation firm? [unverified — confirm] (3) Would a French SEACURE-prime (Thales/Naval Group) ever buy a sovereign COP module from outside its own stack, or build in-house? [unverified] (4) French "Diffusion Restreinte" accreditation cost + timeline for a foreign firm [TBD]. (5) Does the FMES/Observatoire ever broker industry-research conversations, or is it purely analytical? [TBD] (6) Can a CUI-watch demo run on representative (non-classified) data to stay inside the free-discovery model? [TBD]. Until (1)–(3) resolve, treat this as a WATCH-LIST / EDF-consortium-back-door row, not an active pursuit.

Sources (dated 2026-06-05):

• Ministerial "Maîtrise des fonds marins" / seabed-warfare strategy (adopted 14 Feb 2022; know/watch/act to 6000m; protects CUI): defense.gouv.fr ; defense.gouv.fr ; archives.defense.gouv.fr [verified]
• Observatoire Maîtrise des Fonds Marins (stood up by DGRIS within FMES Toulon; operational by 24 Apr 2026; RAdm (ret.) Jacques Mallard; assesses CUI risk; Calliope mission): navalnews.com [verified]
• France-2030 / DGA+SGPI order (27 Sep 2024) of an AUV (Exail A6K-M) + ROV (Louis Dreyfus TravOcean) to 6000m, first French Navy deep-sea capability from ~2026; broader France-2030 seabed-exploration/mastery programme ~€300m o.o.m., of which a ~€100m first milestone funds this initial GISMER drone + ROV: defense.gouv.fr ; opex360.com [verified]
• GISMER / CEPHISMER (operators; CEPHISMER re-org 10 Jan 2023; Toulon/Brest): meretmarine.com ; imagesdefense.gouv.fr [verified]
• Pôle d'excellence grands fonds marins (launched at Forum Innovation Défense, Dec 2025): meretmarine.com [verified]
• Offensive ("volet offensif") component confirmed (Mar 2025) — the decline boundary to avoid: navalnews.com ; opex360.com [verified]
• SEACURE (Thales-led EDF seabed/ASW system-of-systems, demo at sea by 2028; Naval Group/Exail/Leonardo/Atlas/Fincantieri etc.): thalesgroup.com ; opex360.com [verified]
• Naval Group DGA autonomous underwater drone demonstrator: naval-group.com [verified]
• Agence de l'Innovation de Défense (AID) Guichet Unique innovation route: defense.gouv.fr [verified channel; non-French eligibility unverified]
• Nord Stream effect (France's MFM strategy predates Nord Stream — unveiled 14 Feb 2022; the Sep-2022 attacks then prompted France to dive deeper / accelerate undersea security): usnews.com [verified]
• Baltic CUI incidents (Estlink 2, 25 Dec 2024; Nov-2024 fibre cuts) + NATO Baltic Sentry (Jan 2025) + Nordic Warden: en.wikipedia.org ; atlanticcouncil.org [verified]
• EDF 2026 (€1bn, closes 29 Sep 2026, ≥3 MS consortia; seabed topic family): defence-industry-space.ec.europa.eu [verified]
• Internal: discovery-playbook.md, formula-worked-examples.md (LOI formula), legal-blockers-register.md, contact-register.md, outreach-drafts.md; depth pattern mirrored from UC-55 / W-11 / W-15 (sub-sea) in nexus-synergy-ei-sovereign-v2-aw-t03.md; sibling EDA/PESCO seabed package e03-eda-pesco-csip-ums.md.

---