---
title: "Public Corrections & Provenance Audit Ledger — Asymmetric Intelligence (a-i.gi)"
canonical: https://a-i.gi/corrections/
source_html: /corrections/index.html
description: "Asymmetric Intelligence — structured intelligence for a disordered world. Six public-interest monitors, eight commercial intelligence products, one source-disciplined methodology."
generated_at: 2026-11-08T08:30:26Z
content_sha256: bf440350b695b2990aa74e1b1ad275bfa66719e721de21631a93f469d737087f
generator: "fe-fleet/md-twin@1.0"
---
# Public Corrections & Provenance Audit Ledger — Asymmetric Intelligence (a-i.gi)

PUBLIC ACCOUNTABILITY · CORRECTIONS & PROVENANCE LEDGER

# Public Corrections & Claim-Challenger Audit Ledger

Named sources, named reviewers, named gaps. When our automated Grounded Claim Challenger ([/benchmark/summary.json](/benchmark/summary.json)), internal second-opinion review, or a reader identifies an inaccurate claim, composite citation, or count discrepancy across any Asymmetric Intelligence surface, the correction is recorded permanently on this page with its prior state, root cause, and verification link.

10,226

Challenger Tests Emitted

Across 498 JID-monitor units

31 / 31

Blocking Refutations Hand-Cured

Cured in consumer PRs prior to release

199

Direct-Source Contrary Witnesses

Attached via ucs_direct_cure

5

Editorial & Specimen Logs

Logged with full before/after diffs

## 1. Post-Publication Specimen, Copy & Benchmark Ledger

Date
Ledger ID
Surface & Record
Prior Published Text / Defect
Correction Applied & Primary Source Witness

2026-10-27
CORR-2026-10-27-01
advennt.io & payments.gi
Launch-Scope Surgical Auto-Cures ( P-P1-5 )
Grounded Claim Challenger launch-scope passes ( 40499694926 & 40560092257 ) identified 39 Advennt and 26 World Payments direct_source_correction refutations with verbatim statutory spans and source positions.
Surgically repaired 36 of 39 Advennt refutations across 31 JIDs ( asym-intel/advennt#993 , #995 ; 7.69% without-cure) and 25 of 26 World Payments refutations across 15 JIDs ( asym-intel/world-payments#207 , #208 ; 3.85% without-cure); 5 pass-1 cured items re-measured in run 40560092257 flipped from refuted to confirmed (0 re-refuted).

2026-10-26
CORR-2026-10-26-01
sentinel.gi
CA-FI-005 (Homepage Specimen)
Homepage claim specimen paired a composite C$176.9M penalty figure with a November 2024 FINTRAC supervisory guidance URL rather than the October 2025 penalty notice.
Replaced homepage specimen quote and provenance link with FINTRAC’s statutory Administrative Monetary Penalty notice against Xeltox Enterprises Ltd. (Cryptomus, 2025-10-16 ) and verified Canadian MSB registration citation ([fintrac-canafe.gc.ca ↗](https://fintrac-canafe.gc.ca/msb-esm/msb-eng)).

2026-10-26
CORR-2026-10-26-02
a-i.gi
Homepage & Commons
Homepage presented 8 commercial cards without stage differentiation, used a 5-point confidence list inconsistent with /methodology/ , and listed AI Governance in both Commercial and Commons grids.
Separated cards with explicit LIVE (166–170 JIDs), LIVE PREVIEW (3 JIDs on artificial-intelligence.gi ), and PRE-PUBLICATION / SPECIFICATION badges; unified vocabulary to Confirmed / Probable / Uncertain / Unverified + T1–T4 ; deduplicated Commons ledger to 6 monitors.

2026-10-24
CORR-2026-10-24-01
Union Estate
6 Commercial Monitors
Grounded Claim Challenger run ucs-estate-2026-10-24 ( 10,226 claims tested) identified 284 refuted cells ( 31 publication-blocking, 199 direct-source contrary quotes, 50 contested).
All 31 publication-blocking cells were hand-corrected against primary statutory text in consumer PRs ( asym-intel/advennt#977 , asym-intel/crypto#97 , asym-intel/financial-integrity#198 ). All 199 non-blocking cells have the challenger’s verbatim contrary quote attached (`ucs_direct_cure`) pending re-interpretation; no post-cure precision is claimed before re-measurement.

2026-10-18
CORR-2026-10-18-01
a-i.gi/benchmark
Pre-v7 Topic Runs
Early pre-v7 single-pass synthesis runs lost 4 of 6 cells against baseline Deep Research on narrative fluency and missed unindexed regulator circulars.
Retained negative pre-v7 run logs on [a-i.gi/benchmark](/benchmark/); engineered two-tier primary discovery and TRACE B1–B6 citation gates before running the v7 evaluation (7 of 8 cells won).

## 2. Per-Consumer Grounded Claim Challenger Breakdown (Run 2026-10-24)

Consumer Monitor
JID Units
Claims Tested
Confirmed
Refuted
Raw Decided Precision
Blocking Cured
Cure PR

advennt.io (Gambling) 123 1,394 505 29 94.6% 6 asym-intel/advennt#977 , #993 , #995

payments.gi (Payments) 41 2,326 1,454 39 97.4% 8 asym-intel/world-payments#197 , #207 , #208

dataprotection.gi (Data Protection) 130 2,441 1,190 76 94.0% 7 asym-intel/data-protection#112

cryptoassets.gi (Cryptoassets) 139 1,877 843 63 93.0% 2 asym-intel/crypto#97

sentinel.gi (Financial Integrity) 62 1,607 437 76 85.2% 8 asym-intel/financial-integrity#198

artificial-intelligence.gi (AI Governance) 3 581 247 1 99.6% 0 asym-intel/artificial-intelligence-gi#158