AI Dispatch — Daily Dispatch, September 14, 2026

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.

## Wire items today

Three items ran. Two of the three entities covered — KHAN and TRam Studio — were promoted from candidate to published in the last 24 hours. The third, MyPolicyReview.ai, **remains a candidate**: today's piece is a follow-up on an entity this outlet first covered September 11, not a promotion, and its coverage below is on that basis.

- **KHAN** — an autonomous trading agent on the Solana blockchain that calls its public ledger "the single source of truth for every action," with each entry "a verified on-chain transaction" covering claims, buys, sells, payouts and governance proposals. That is a settlement path this outlet's rubric credits toward `outcome_settlement`. But the same page stops short of the stronger claim on the record itself: it does not describe the ledger as tamper-evident, hash-chained or customer-reviewable — the properties `audit_trail` scores highest. The page discloses a realized loss as readily as a gain and states "no returns are promised on any position." `insurance_indemnity`, `compliance_certs` and `runtime_governance` are all `no_public_information` on its record. - **TRam Studio** — a workflow-automation platform built on Microsoft Azure that markets "Every run logged, step by step" and "Hard spend limits, no surprise bills" — `present` evidence toward `audit_trail` and `runtime_governance` both. Its published commercial terms are a flat "Fixed price starting at $5,000" implementation pilot, not the per-task dollar costs shown in its own marketing screenshots, which are vendor depictions rather than documented customer runs. `compliance_certs` and `insurance_available` are not published on the page reviewed. - **MyPolicyReview.ai** (candidate, not published) — a fuller read of the AI policy-review vendor's site turned up a lender-facing "Compliance dashboard with audit export" that this outlet's September 11 article, drawn from an earlier pass over the same page, had reported as absent. The finding narrows, not closes, the gap: the feature is scoped to lenders rather than individual policyholders, and compliance certification, insurance coverage of the vendor's own service, a liability cap and indemnification terms remain unpublished.

## What changed in the index this week

Queried directly against the live database at draft time:

- **Published entities: 792** (762 agent, 30 vendor). Candidates: 2,311 (2,280 agent, 31 vendor). Archived: 2. - **15 candidate-to-published promotions, 5 published-to-candidate demotions** in the trailing seven days (net +10). In the last 24 hours alone: 5 promotions (Orchestra Ads, EmAGI, TRam Studio, KHAN, IELTS Writing Practice — three of the five not covered as standalone wire items today), 1 demotion. - **4,778 `entity_fields` rows touched** in the trailing seven days — 945 new or corrected `present` values, 3,833 flipped to `no_public_information`. - **2,040 `change_queue` items resolved against 1,400 newly detected** in the same seven days — the pending backlog fell to 2,533, down from 4,528 reported in the previous issue. This is the first week on record where resolution has outpaced new detections rather than the reverse. - Full five-dimension scorecard coverage, human-scored dimensions only: **96 of 792 published entities (12.1%)**.

## One assurance/insurance development

Human-scored dimensions only — the automated scorer running alongside the Fact-Checker cannot currently write a 0 or a 3, so blending automated and human scores would overstate coverage of the gap the rubric exists to measure:

- **`insurance_indemnity`: 99 of 100 human-scored published entities (99.0%) sit at 0 or 1**, 96 of those a flat zero. None of today's three wire entities carries a published insurance or indemnity term of its own, KHAN and TRam Studio included, despite one settling by verified on-chain transaction and the other capping runtime spend. - `outcome_settlement`: 93 of 96 (96.9%) at 0-1, 91 at a flat zero. KHAN's on-chain settlement is the kind of mechanism this dimension is built to reward, though the entity has not yet been hand-scored on it. - `audit_trail`: 84 of 97 (86.6%) at 0-1, 63 at a flat zero. Today's wire items split on this dimension: TRam Studio's per-run log and MyPolicyReview.ai's newly surfaced audit export are both `present` evidence, while KHAN's page stops at "verified," not tamper-evident. - `compliance`: 80 of 97 (82.5%) at 0-1, for context, not counted toward the standing three. - `runtime_governance`: 69 of 97 (71.1%) at 0-1, same caveat.

The competitor wedge also holds: across theaiagentindex.com (378 listed, 0 featured) and aiagentsdirectory.com (3,035 listed, 15 featured — up from 14 in the last issue) — 3,413 combined — zero name any of the insurance or liability carriers we track (Armilla, Munich Re aiSure, Lloyd's, Chaucer, Embroker, Mosaic, Relm, Testudo, Coalition, Vouch). Neither 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.*