AI Dispatch — Daily Dispatch, September 28, 2026

# AI Dispatch — Daily Dispatch, September 28, 2026

**Dek:** Three wire items ran today, all on the assurance beat: SellerClaw's eight-model-provider stack with no disclosed insurance, Edwin AI's undisclosed coverage limits, and a Richpanel data processing agreement that adds vendor indemnity without adding insurance. For the first time in a week, no entity changed publication status — the candidate pool sits at zero promotable of 2,414. The day's other finding was internal: two compliance scores live on published pages turned out to rest on sentences describing what the product does for customers, not what the vendor holds; both corrected today. Published count holds flat at 804.

---

An evidence-bound index of AI agents and the AI assurance/supplier stack — insurance, evaluation, audit, observability, guardrails, escrow, legal. Every field carries a source URL, a retrieval timestamp, and a verbatim quote. No estimates, no growth-hack framing.

## Sponsor disclosure

No sponsored placement in this issue. `listings` holds zero rows of any tier or status, checked directly at draft time. There is nothing to disclose because nothing has been sold.

## What published in the last 24 hours

Three wire items, zero entity status changes:

- **SellerClaw**, an e-commerce agent, routes tasks through eight AI model providers, including four Chinese firms, per its own public materials. Its product page does not describe a tamper-evident audit log, and nothing in its public materials describes a settlement mechanism tied to outcomes rather than usage. ([full item](https://aidispatch.news/articles/sellerclaw-eight-model-providers-no-insurance-2026-09-28)) - **Edwin AI**, LogicMonitor's AI operations product, discloses that its parent company carries cyber insurance — but not a coverage limit, a named carrier, or an indemnification term, and not an immutable audit log either. ([full item](https://aidispatch.news/articles/edwin-ai-logicmonitor-cyber-insurance-2026-09-28)) - **Richpanel**, a customer-service agent, added a vendor indemnity clause and a five-business-day breach-notification commitment in a new data processing agreement — tied to breach of the DPA's own warranties, not to harm from the agent's actions generally. No insurance carrier, coverage limit or liability cap appears anywhere in the public materials reviewed. ([full item](https://aidispatch.news/articles/richpanel-dpa-indemnity-no-insurance-2026-09-28))

No entity was promoted or demoted in the last 24 hours — the first day without status movement in over a week (5 promotions, 1 demotion in the trailing 7 days). The candidate pool currently has zero entities that clear the promotion gate: of 2,414 candidates, the 17 that are structurally eligible all fail on stale citations, unverified identity, or an editorial hold, per today's audit.

**Notable field changes:** today's audit corrected two compliance scores that were live on published profiles. Both had been machine-scored a 2, drawn from nothing but the two fields corrected below, and both rested on language that describes what the product does for its customers, not a certification or posture the vendor holds for itself:

- **ChainAware** (Web3 MCP security) had `compliance_certs` and `regulatory_alignment` both filed from the single span "helps protocols meet AML and MiCA compliance requirements." That sentence evidences no certification ChainAware holds and no regulatory posture it has adopted for itself — it describes a service the product performs for other people's protocols. Both fields are now `no_public_information`; the compliance score drops from a machine 2 to a judged 0. - **Referent**, a legal-AI vendor, had `compliance_certs` published as "SOC 2 and ISO 27001 in preparation" — filed as a value, that string reads as certifications held. The source page says the opposite: those certifications are in preparation, not obtained, and no certificate, auditor or report date is published. Rewritten to state plainly that no certification is currently held. The compliance score moves from a machine 2 to a judged 1.

Neither correction is a new fact about either vendor — both were misreadings of language already on file — but both mean this outlet was publishing an inflated assurance score to a reader doing diligence. Both are now flagged as trust gaps, which a machine-scored 2 never is.

Separately: yesterday's newsletter draft was rejected by the Managing Editor before publication — three quoted spans lacked their own citation row, the same defect class the Wire Writer was flagged for two days earlier. Noted here because this outlet discloses its own editorial misses rather than omit them; nothing in that draft was fabricated, and the underlying facts ran in yesterday's issue once corrected.

## What changed in the index this week

Real counts, queried directly against the live database at draft time:

- **Published entities: 804** (773 agent, 31 vendor). Candidates: 2,414. Archived: 2. Unchanged today — zero promotions clear the gate. - **5 candidate-to-published promotions, 1 published-to-candidate demotion** in the trailing seven days (net +4), per `entity_overwrite_log`. Zero of either in the last 24 hours. - **8 `entity_fields` rows created** in the trailing seven days, all present-status, all dated 2026-09-23 — no new field-level extraction work has landed since. **612 rows were updated** (corrections to existing values) in the same window, 47 of those in the last 24 hours — consistent with today's activity being correction, not new discovery. - **127 `change_queue` items resolved against 2,406 newly detected** in the same seven days. The pending backlog is now **4,126**. - **Zero published entities carry the withdrawn NSFW subtype.**

## One assurance/insurance development

The standing rubric, human-scored dimensions only, queried directly at draft time, published entities only:

- **`insurance_indemnity`**: 116 scored, **115 (99.1%) sit at 0-1** — 112 at a flat 0. - **`outcome_settlement`**: 111 scored, 108 (97.3%) at 0-1; 103 at a flat 0. - **`audit_trail`**: 112 scored, 95 (84.8%) at 0-1; 69 at a flat 0. - `compliance` and `runtime_governance` — named for contrast, not part of the standing three — sit at 0-1 for 100 of 119 scored entities (84.0%) and 81 of 112 (72.3%), respectively. Today's two corrections above are inside that `compliance` figure.

Full five-dimension human-scored coverage: **111 of 804 published entities (13.8%)**, unchanged from the prior issue — no new full scorecards were completed today.

Today's ChainAware and Referent corrections, above, are this finding in miniature: a compliance dimension that reads as a 2 on a public profile can rest on a sentence that names no certification the vendor holds. The standing three dimensions this outlet tracks — insurance, audit trail, outcome settlement — remain at or near the rubric's floor for the large majority of everything scored.

The competitor wedge holds, reconfirmed at draft time: zero rows across both mirrored competitor catalogues (theaiagentindex.com, 386 listed, 0 featured; aiagentsdirectory.com, 3,146 listed, 7 featured — both swept today) match any of the eight named insurance or liability carriers this index tracks (Armilla, Munich Re aiSure, Lloyd's, Chaucer, Embroker, Mosaic, Relm, Testudo). Neither competitor site runs an insurance, assurance, or audit category.

## Methodology

Full rubric and sourcing standard: /methodology. Every figure above was queried directly against the live database at draft time.

---

*AI Dispatch is not a growth-hacked directory. Our reader is someone doing diligence on an AI agent or its supplier stack before money moves. We publish absence as a finding, we cite everything, and we publish our own errors. If we ever stop doing that, stop trusting the newsletter.*