# Cross-Patent Demarcation Memo — Band A (Wave 1)

> **CONFIDENTIAL — Patent Work Product. Attorney-Client Privileged. Not Filed. Internal Use Only.**
> Prepared: 2026-05-30 | Author: Verixa Patent Attorney workbench | Resolves: OI-2 (Decision 2)
> Status: **attorney-review-ready** (preliminary on element-level claim confirmation — see §Caveat)

---

## Purpose

The current demarcation memo on file covers only **P02 ↔ P03**. This memo extends documented
demarcation to **all 8 cross-patent pairs** flagged in `00_Portfolio_PreFiling_Status_Memo_2026-05-29.md`
(Decision 2 / OI-2). It exists to:

1. **Unblock P02 and P02b** (both on 🟡 HOLD pending P02↔P02b and P01↔P02b demarcation).
2. Document double-patenting / overlapping-scope boundaries before the complete-specification stage
   (FD+12mo, **2027-06-15**), so each invention claims a distinct technical mechanism.
3. Frame the one strategy decision that cannot be deferred for **P04 ↔ P10**.

**Stage rule.** Demarcation is principally a **complete-specification** discipline. A provisional
secures priority for *what it discloses*, not for claim language. Therefore: disclosure-rich
provisionals filed by 2026-06-15 are **not blocked** by claim-scope overlap. The HOLD on P02/P02b is
a *strategic sequencing* hold (lead-element framing + applicant identity), not a statutory bar to
provisional filing. This memo lets P02/P02b move from HOLD → READY for provisional purposes while
preserving the demarcation work for complete-spec.

---

## Band A scope under review

| Patent | Short title | Core component(s) | Independent-claim spine (verified in Form 2) |
|---|---|---|---|
| **P01** | Platform Validated Model Lockstep | H1/H2/H3 + E1 | Validated model/version lockstep across the AI processing pipeline; §6.11 + §10 composition |
| **P02** | Cryptographic Per-Inference AI Governance | H1/H2/H3 | Per-inference audit **record chain** + model-version governance + agent qualification |
| **P02b** | AI Output Segregation Engine | E2 | Output-type **classification** + prohibited-field **write-prevention gate** + sanitizer + DQG staging + advisory evidence pack |
| **P03** | Cryptographic Hash-Chain Audit Log | E3/E4 | **Single-tenant** per-tenant hash chain + TDAL/RLS + append-only trigger |
| **P04** | HITL E-Sig Ceremony | E6/E7 | Human-in-the-loop e-signature **ceremony** + state-verified workflow transition |
| **P05** | Pre-Execution Context Gate | E5 | Per-tenant role-scoped **dimension profiles** + home-site sentinel + scope enforcement |
| **P09** | Cross-Tenant Hash-Chain Replay Resistance | E4a + E3 | **Cross-tenant** boundary enforcement + replay resistance + advisory lock |
| **P10** | Domain-Weighted Agent Deliberation | E7 | GxP-event-type routing + **domain-relevance matrix** + conflict adjudication + escalation determination |

*(P06 — Dual Stack Composition — carries no standalone Form 2; covered by P01 §6.11 + §10. Excluded from the pair analysis; its composition claim is drafted at complete-spec.)*

---

## The 8 demarcation pairs

For each pair: **Overlap zone** (where claims could collide) → **Demarcation line** (what each patent
owns) → **Action** (claim-scoping at complete-spec) → **Residual risk**.

### Pair 1 — P01 ↔ P02b  *(unblocks P02b — HIGH)*
- **Overlap zone:** P01 §6.1.4 (AI gateway / context-aware request handling) vs P02b E2 (per-inference output-type classification + write-prevention).
- **Demarcation line:** **P01 owns the audit/lockstep chain** — model-version validation and the processing pipeline through which inference requests flow. **P02b owns what happens to the *output*** — classifying output type and **preventing writes to prohibited GxP fields** before persistence, plus the forbidden-term sanitizer, DQG staging, and advisory evidence pack.
- **Action:** In the P02b complete-spec, claim the **classification + prohibited-field gate** as the inventive core; reference (do not re-claim) the P01 pipeline as the environment. In P01, recite the gateway as a *transport/validation* mechanism, not as an output-segregation mechanism.
- **Residual risk:** Low once the "pipeline vs output-gate" split is recited. Confirm P01 §6.1.4 language does not independently claim output suppression — *Preliminary; element-level confirmation required against P01 §6.1.4.*

### Pair 2 — P02 ↔ P02b  *(unblocks both — HIGH)*
- **Overlap zone:** Both operate "per inference." P02 = per-inference **audit record chain**; P02b = per-inference **output-type classification + prohibited-field gate**.
- **Demarcation line:** **P02 owns the immutable per-inference *audit record*** (what was run, which model/version, qualification state) and its chaining. **P02b owns the *pre-persistence decision*** — classify the output and block prohibited-field writes **before** the audit record is written.
- **Action:** Sequence the claims temporally: P02b gate **precedes** P02 record creation. P02b claims "determine output type → prevent prohibited write → emit advisory evidence"; P02 claims "record the (permitted) inference event into a per-inference chain with model-version + qualification metadata." Avoid any P02 claim element that performs classification, and any P02b element that claims the audit *chain* itself.
- **Residual risk:** Medium — these are the two most adjacent patents. The temporal/functional split (gate-then-record) is the cleanest non-overlapping boundary. Recommend a single shared §3 paragraph in both specs stating the boundary explicitly.

### Pair 3 — P02b §3 Summary lead-element restructure *(follows Pair 1 — MEDIUM; = OI-4)*
- **Not a patent-vs-patent pair**, but gated by Pair 1. P02b §3 currently leads with (a) classifier and (b) prohibited-field prevention — the two elements **most exposed to India §3(k)** if not anchored to a technical effect.
- **Action:** Once the P01↔P02b line (Pair 1) is fixed, restructure P02b §3 to **lead with (c) forbidden-term sanitizer, (e) DQG staging structure, (g) advisory evidence pack** (more differentiated, more technical), and demote (a)(b) to dependent positions. This is editorial/strategic, not a scope change — route through the Style Editor, **scope-change = NO**.

### Pair 4 — P01 ↔ P05  *(MEDIUM)*
- **Overlap zone:** P01 §6.1.4 context-aware requests vs P05 context-gate **profiles**.
- **Demarcation line:** **P01 owns the processing pipeline** (how a request is validated/handled). **P05 owns the *profile store + RLS + scope resolution*** — per-tenant role-scoped dimension profiles, home-site sentinel resolution, scope enforcement *before* execution.
- **Action:** P05 claims the **pre-execution gate data structures and resolution algorithm**; P01 references "a context gate (see P05)" as an input condition without claiming the profile store. P05 already source-verified against `context-gate/service.ts`, `routes.ts`, `audit-registry.ts`, migrations 092/132/259 — anchor claims there.
- **Residual risk:** Low. Distinct modules and distinct inventive mechanism.

### Pair 5 — P01 ↔ P04  *(MEDIUM)*
- **Overlap zone:** P01 §6.6 (HITL integration into the pipeline) vs P04 E6/E7 (the e-signature **ceremony** itself).
- **Demarcation line:** **P01 owns pipeline integration** (the pipeline pauses for human approval). **P04 owns the *ceremony structure*** — the e-signature meaning binding, state-verified workflow transition, and the decision lifecycle.
- **Action:** P04 claims the ceremony mechanics (esig.service / hitl.service / workflow-engine state verification); P01 recites "a human-approval checkpoint" as a pipeline state without claiming the signature ceremony. 
- **Residual risk:** Low.

### Pair 6 — P03 ↔ P09  *(MEDIUM)*
- **Overlap zone:** Both recite a per-tenant cryptographic hash chain over the audit log + TDAL/RLS (P09 Claim 1 elements 1–3 mirror P03's E3/E4 spine).
- **Demarcation line:** **P03 owns the single-tenant chain** — per-tenant hash chaining, append-only enforcement, intra-tenant tamper-evidence. **P09 owns the *cross-tenant boundary layer*** — cross-tenant replay resistance, the advisory lock, and boundary enforcement that P03 does not claim.
- **Action:** P09 must claim the **cross-tenant** delta as the inventive step over P03 (replay resistance + advisory lock + boundary enforcement), explicitly positioning P03's single-tenant chain as the base on which P09 builds. P03 claims must **not** recite cross-tenant replay resistance.
- **Residual risk:** Medium — P09 Claim 1 currently re-recites the GUC/RLS/hash-chain base (elements 1–3) shared with P03. At complete-spec, restructure P09 Claim 1 so the base is a preamble/dependent reference and the **cross-tenant replay-resistance + advisory-lock** elements are the independent inventive core. *Preliminary; confirm against P03 independent claim once P03 claims are finalized.*

### Pair 7 — P02 ↔ P04  *(LOW–MEDIUM)*
- **Overlap zone:** P02 model-version pinning/governance vs P04 HITL ceremony — both touch "who/what is authorized to act."
- **Demarcation line:** **P02 owns *model* governance** (version pinning, agent qualification of the AI). **P04 owns *human* approval** (the e-signature ceremony by a person).
- **Action:** Keep machine-qualification (P02) and human-approval (P04) as separate authorization layers. No claim in P02 should recite human e-signature; no claim in P04 should recite model-version pinning.
- **Residual risk:** Low.

### Pair 8 — P04 ↔ P10  *(🔴 HIGHEST RISK — strategy decision required)*
- **Overlap zone:** **Confirmed by spec inspection.** The P04 Form 2 itself describes Component E7 modules — `deliberation.service.ts`, `composite-scoring.service.ts`, `agent-registry.service.ts`, `panel-session.service.ts` (four-phase panel deliberation, domain-relevance-weighted scoring, agent registry with per-event-type weights). P10's independent claims (GxP-event-type routing, **domain-relevance matrix**, conflict adjudication, escalation determination) claim **the same E7 mechanism**. This is direct claim-scope collision, not adjacency.
- **Demarcation line (two viable options):**
  - **Option A — P10 independent:** Strip E7 deliberation/composite-scoring from **P04**'s claimed scope (P04 = E6 e-signature ceremony + workflow transition only; E7 referenced as environment). P10 stands as the independent owner of domain-weighted deliberation. *Cleanest if deliberation is a distinct commercial/inventive asset worth its own patent family.*
  - **Option B — P10 as P04 divisional sub-claim:** Keep E7 within the P04 family; P10 becomes a divisional/dependent at complete-spec. *Cleaner inventorship/priority story; one family; lower filing count, but P10 loses independent standing.*
- **Action / decision required from Vimal:** **Choose Option A or B** before the complete-spec stage (2027-06-15). **Provisional filing on 2026-06-15 is NOT blocked** — file both P04 and P10 provisionals disclosure-rich; the election is a complete-spec decision.

#### ★ DECISION (recorded 2026-05-30) — ELECT OPTION A (P10 independent), CONDITIONED

**Decision:** Elect **Option A** — P10 is the independent owner of the E7 domain-weighted deliberation mechanism.

**Condition (non-negotiable for Option A to be valid):** At the **complete-specification** stage, **P04 must be rewritten to claim only Component E6 (cryptographic HITL e-signature ceremony) + Component E3 (per-tenant RLS isolation substrate) + the E6↔E7 handover bridge as *environment***. **P04 must DROP all claims to E7's internal deliberation mechanism** (GxP-event-type routing, domain-relevance matrix, four-phase deliberation, three-type conflict detection, escalation-level determination). Two independent applications both *claiming* E7 = fatal same-invention double patenting. P10 claims E7's internal mechanism exclusively; P04 may *describe* E7 as the source of the DRT it signs, but must not claim it.

**Spec-grounded basis for the decision:** P04's Form 2 currently frames E7 as a **co-inventive component** — Field of Invention limb (b), Summary ("two inventive components integrated through a structured handover bridge"), Technical Effects (vi), Component E7 §4, Advantages (e), and the Abstract all claim the multi-agent weighted deliberation system as part of the P04 invention. This is genuine double-claiming exposure, not adjacency. P10's independent claims (routing + qualification gate + domain-relevance matrix) claim the same mechanism.

**Why Option A (rationale of record):**
1. **Distinct inventive concepts** — E7 (adjudicating conflicting multi-agent outputs) and E6 (binding a human signature) solve different technical problems, face different prior art, and carry different eligibility postures. Bundling weakens both.
2. **Unity-of-invention hygiene** — P04 currently bundles E3+E6+E7; splitting E7 into P10 removes a lack-of-unity exposure at complete-spec/PCT and gives each application a clean single-inventive-concept story.
3. **§3(k) risk isolation (decisive for India)** — E7 is the element most exposed to a §3(k) "algorithm / computer programme *per se*" objection. Independent P10 *contains* that risk; a §3(k) hit on the deliberation algorithm then cannot entangle P04's e-signature ceremony (which has the stronger technical-effect story: cryptographic chaining, advisory locks, live authority re-validation, state-bypass prevention).
4. **Separable commercial value** — domain-weighted deliberation is a standalone, separately-licensable "panel IP" asset; independent standing preserves the ability to assert/license/sell it alone.
5. **Zero priority cost** — both provisionals already disclose E7 richly as of 2026-06-15, so P10 keeps the priority date whether independent or divisional. Nothing is lost by electing A now.

**Fallback / optionality:** The *direction* is locked as A now (so P04 and P10 are disclosed for independence). The *execution* occurs at complete-spec (2027-06-15). If E7 draws a §3(k) objection during prosecution, A→B (Section 16 divisional of P04) remains an inexpensive conversion; the reverse is not. Therefore A preserves maximum optionality.

**Decision owner / status:** Recommended by Verixa Patent Attorney workbench; **adopted by Vimal (Founder) 2026-05-30**. Final claim execution and double-patenting clearance reserved to outside Indian patent counsel at complete-spec.

- **Residual risk:** **Low once the condition is honored at complete-spec** (P04 drops E7 claims; P10 claims E7 exclusively). **High if the condition is not honored** — if both applications claim E7, double-patenting rejection is likely.

---

## Summary matrix

| Pair | Risk | Boundary (who owns what) | Decision stage | Blocks provisional? |
|---|---|---|---|---|
| P01 ↔ P02b | High | P01 = pipeline/audit chain · P02b = output classify + write-gate | Complete-spec | No |
| P02 ↔ P02b | High | P02 = audit record chain · P02b = pre-write gate (gate-then-record) | Complete-spec | No |
| P02b §3 restructure | Med | Lead with (c)(e)(g); demote (a)(b) | Now (editorial) | No |
| P01 ↔ P05 | Med | P01 = pipeline · P05 = profile store + scope resolution | Complete-spec | No |
| P01 ↔ P04 | Med | P01 = pipeline integration · P04 = ceremony structure | Complete-spec | No |
| P03 ↔ P09 | Med | P03 = single-tenant chain · P09 = cross-tenant + replay + advisory lock | Complete-spec | No |
| P02 ↔ P04 | Low | P02 = model governance · P04 = human approval | Complete-spec | No |
| **P04 ↔ P10** | **High → Low (conditioned)** | **ELECTED Option A: P10 owns E7; P04 drops E7 claims at complete-spec** | **Direction locked 2026-05-30; execute at complete-spec** | **No** |

**Net effect:** With demarcation lines documented above, **P02 and P02b move HOLD → READY for provisional filing** (subject only to the portfolio-wide applicant-identity decision). All eight boundaries are complete-spec execution items; **none blocks the 2026-06-15 provisional deadline.**

---

## One decision that cannot be deferred — RESOLVED 2026-05-30

**P04 ↔ P10 election: ELECTED Option A (P10 independent), conditioned.** Direction locked 2026-05-30 (see
Pair 8 ★ DECISION). Both 2026-06-15 provisionals are filed disclosure-rich for independence. **Carry-forward
action to complete-spec (2027-06-15):** P04 complete-spec must claim **only E6 + E3 + handover-bridge-as-environment
and DROP all E7 claims**; P10 claims E7's deliberation mechanism exclusively. Fallback A→B (Section 16 divisional)
remains available if E7 draws a §3(k) objection in prosecution.

---

## Model Selection Note
- Model used: claude-opus-4-8
- Effort level: high
- Why selected: cross-patent claim-scope demarcation grounded in on-disk Form 2 specs; multi-document reconciliation
- Output status: **attorney-review-ready** for the demarcation framework; **preliminary** on element-level claim confirmation (Pairs 1 and 6 carry explicit confirmation flags)
- Stronger model review recommended: yes, at complete-spec claim-drafting stage (xhigh + outside counsel)
- Reason: provisional-stage demarcation is directional; final double-patenting clearance is a claim-construction determination for counsel

## Source Ledger
| Source type | Source path | Date | Claim supported | Limitation |
|---|---|---|---|---|
| Portfolio status | `00_Portfolio_PreFiling_Status_Memo_2026-05-29.md` | 2026-05-29 | 8-pair list, overlap risks, Decisions 1–3 | Drafter work product |
| Form 2 spec | P02b `VeriXA_P02b_Form2_Provisional_Specification.md` | on-disk 2026-05-30 | P02b classify/HITL/write-gate spine | Provisional |
| Form 2 spec | P03 `VeriXA_E4_Form2_Provisional_Specification.md` | on-disk 2026-05-30 | P03 single-tenant chain + TDAL/RLS | Provisional |
| Form 2 spec | P04 `VeriXA_E6E7_Form2_Provisional_Specification.md` | on-disk 2026-05-30 | P04↔P10 overlap (E7 modules in P04 spec) | Provisional |
| Form 2 spec | P09 `Verixa_Form2_Provisional_Specification.md` | on-disk 2026-05-30 | P09 Claim 1 elements 1–3 (cross-tenant) | Provisional |
| Form 2 spec | P10 `Verixa_Form2_Provisional_Specification.md` | on-disk 2026-05-30 | P10 routing + domain-relevance matrix | Provisional |

> This is a Verixa patent attorney workbench output, not a final legal opinion. Qualified patent counsel
> and local counsel where required must verify law, facts, repo evidence, claim scope, filing authority,
> prosecution strategy, FTO conclusions, trade-secret risk, publication clearance, and jurisdiction-specific
> requirements before action. Element-level double-patenting clearance is a claim-construction determination
> reserved to counsel at the complete-specification stage.