GAUNTLET — Payment Screening · Bulwark Compliance

L2 — Pending Review

—

payments in queue

Priority

—

high-confidence / BIC

L1 — Auto-Cleared (Rules)

0

payments auto-cleared

Screened Today

—

transactions

Avg Match Score

—

across pending payments

L2 Analyst Queue — Pending Review

Tier	Severity	Payment Ref	Status	Fmt	Originator	Beneficiary	Best List Match	Score	Hits	Amount	Age / SLA

Open Records

—

blocks + rejects

Report Due ≤ 3 Days

—

OFAC 10-business-day clock

Overdue

—

report past due

ARBP Scope

—

annual report of blocked property

Blocked Property & Reject Reports — OFAC report due T+10 business days (weekends skipped; federal holidays not)

Type	Payment Ref	DC	Target	SDN UID	Amount	Event	Report Due	Days	ARBP	Status

Automated Triage Performance — 24h

last 24h

Total Screened

—

Auto-Cleared

—

Rules-Assisted → Analyst

—

Pending Review

—

Disposition Codes — Last 24h

Loading…

OFAC List Currency

SDN List—

Consolidated—

Live publish dates from the reference plane (/health/data).

Rule Proposals — Patterns Discovered from Disposition Data

Analyst Consistency Monitor — February 2026

Regulatory Change Impact Simulator

Screening-Rules Change Log — thresholds · review chain / SLA · screening policy · allowlist & list governance · settings proposals

Every configuration change to the screening model, read from the append-only audit log. This is the change-control evidence an examiner reviews (docs/model-governance.md).

Loading…

Allowlist Governance — expiry · review-by · justification sweep

Active entries on allowlist-action lists. An allowlist suppresses false positives, so each entry carries an expiry date, a review-by date, and a justification. Entries past their review date (or expired, or missing a justification) need action in Internal Lists — request a modify or removal there. An allowlist never suppresses a true SDN match.

Loading…

LLM Adjudication — verdict vs human-disposition agreement (shadow-mode measurement)

Phase 2 of the L1 program (docs/model-governance.md §11): an LLM adjudicates held reviews in the background (shadow → propose → delegated, four-eyes-controlled per tenant). This panel is the promotion evidence — weekly agreement between the LLM verdict and the eventual human disposition, split by the in-band class the deterministic tier cannot discriminate, plus the FP-verdict precision that gates any move to delegated. Counts only; per-review verdicts live on the payment record (propose stage and above).

Loading…

Open RFIs — review SLA paused while outstanding

Paste an inbound MT199 response

With automatic inbound enabled (Admin → RFI Channels), gateway files and mailbox replies are polled and threaded for you — pasting here is the manual fallback. A response quoting our RFI reference in its field 21 threads onto the RFI automatically and closes it; anything else lands in the queue below for manual attach/dismiss.

Unmatched inbound queue

HSI Escalations — open

OFAC List Search — Live Database (19,154 entries)

Enter a search term to query the live OFAC SDN/Consolidated database.

Internal Lists

Loading…

Internal Lists

Client-supplied screening lists — exited customers, internal SAR subjects, do-not-bank, PEP overlays, allowlists. They screen on the same engine and 0–100 scale as OFAC; entries are four-eyes (added pending, activated by a second reviewer); an allowlist never suppresses a true SDN match. Select a list or create one.

⚑

Circumvention Log

Historical record of all detected circumvention patterns. Active detection surfaces inline during alert review.

Circumvention Alerts

Total Open

0

Blocks

0

Rejects

0

Unblocks

0

ARBP

0

T+10 At Risk

0

AEGIS Cases

Case ID	Type	Status	Entity	Program	Amount	T+10	Owner
Loading cases from database...

Trans#	BDR/CIF	Target	Program	Date Blocked	Original	Current (X-One)	Interest	ARBP	Status
Loading...

Voluntary Self-Disclosure Cases

Reporting & Management Information

Monitored Subjects

Loading…

Risk-Shift Events

Loading…

Sweep Runs

Scheduled & list-change sweeps, manual rescreens, and one-time onboarding backtests.

Loading…

Coverage Freshness (SLA)

Screened since SLA (hrs)

Pick a reference time to compute coverage.

Attestation Ledger

Loading…

ISO 20022 / SWIFT Message Format Mappings

⚠

Full Escalation Mode

All filtering rules are disabled. Every screening alert is routed to human review. Activate rules individually below after team agreement.

Tenant Screening Threshold

The match score at or above which a screened name becomes an alert. This is the tenant-wide default — distinct from the per-disposition auto-close thresholds below.

Loading…

Per-Disposition Scoring Thresholds

Code	Min Confidence	Sampling Rate	Override Threshold	Current Override Rate	Active

Model Version Registry

Rule Change History

General License Inventory

Specific Licenses (Institution-Held)

No specific licenses registered. HSI can add OFAC-assigned specific licenses here.

Audit Log

Immutable record of all system actions and dispositions. Showing the most recent 200 matching entries.

When	User	Action	Entity	Detail

Notifications

Four-Eyes Settings Change Control — threshold · review chain · auto-release · SLA

Model-risk change control (docs/model-governance.md): one approver proposes a settings change, a different admin approves before it takes effect. The database refuses self-approval. One pending proposal per area at a time.

Loading…

Propose: match threshold

New tenant default threshold (0–1; blank = clear to deployment default)

Rationale

Propose: review chain / auto-release / SLA

Mode

Automated (rules-based) L1 auto-release

Auto-release ceiling (0–0.60 · hard cap)

QC sample rate (0–1 · default 0.20)

Review SLA (hours)

LLM adjudication stage (staged rollout)

LLM QC sample rate (0–1 · default 1.00)

Rationale

Blank fields propose the product default for that value (the proposal always carries the full set — what you see is exactly what takes effect on approval).

Proposal history

RFI dispatch channels — SWIFT gateway (RJE over SFTP) + email (SMTP)

Loading…

Invite a teammate

Pre-provision list rights by email. Sign-in happens through the Bulwark hub — accounts are created there; this only sets what they can do in Gauntlet when they first sign in.

Teammate email

Can request Can approve Senior (L3) Admin

Access Rights

Set who in your organization may request and approve internal-list changes.

Loading…

Monitored-List Sources

Loading…

Screening Policy — which lists screen which record types

A cell toggles whether that sanctions list screens that record type. Defaults: OFAC on, EU/UK/UN off. Turning OFAC off for a record type removes US-required sanctions screening on that flow — allowed, but logged and warned. Rows with no explicit setting use the default.

Loading…

Review Chain — L1 (automated triage, rules-based) · L2 (human) · L3 (senior)

Defaults reproduce today's behavior: two-human maker/checker on every terminal disposition, deterministic rules-based (L1) auto-release of clear false positives below the ceiling, and a deterministic QC sample of auto-releases into the L2 queue (default 20%). L1 is rules-based triage — no machine-learning or LLM judgment is involved. The ceiling is hard-capped at 0.60 — a conservative operating limit supported by mechanism-level testing and corpus-based validation (see docs/model-governance.md for validation status and limitations).

Chain mode

Two-human (default) — every terminal disposition is maker/checker four-eyes; funds move only on the second analyst's approval. One-human — the automated (L1) rules-based recommendation is the documented first eye; the single L2 decision executes immediately, uniformly across release, reject and block. Examiner consideration: this replaces the human second eye with the recorded automated (rules-based) leg — document the rationale in your compliance program before enabling.

L1 auto-release (rules-based) — release clear false positives automatically. Off = L1 only RECOMMENDS release; every payment waits for a human.

Auto-release ceiling (0–0.60] — hard cap

QC sample rate [0–1] (default 0.20)

Review SLA (hours; default 24)

Screening-Queue Health — dual lanes

Loading…

Issue a partner key

Machine-to-machine keys for /api/partner/* (screening, batch submission, roster ingest, result polls). The key is shown once at creation — only its hash is stored.

Key name *

screen ingest results batch

Partner Keys

Loading…