AI Dispatch — Daily Dispatch, August 25, 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. The commercial layer has not been built; there is nothing to disclose because nothing has been sold.

## Published today

No new entities were promoted or demoted in the last 24 hours — a quiet day on the index itself after yesterday's batch of eight. Three pieces published:

- **The Trust Gap, Vol. 4** — the Analyst's own weekly report caught and corrected its prior volume before publication: last week's "absence rate improved" read was a denominator-change artifact (real change: -0.18pp, not the +4.5pp the earlier draft would have shown), and a hand-read of all 8 published `liability_cap` fields found 7 of 8 (87.5%) misfiled into the wrong rubric field. Published anyway, against our own favor. - **Stripe's Agentic Commerce Protocol** lets AI agents "help you earn and spend funds" on a platform processing "$1.9T in payments volume" in 2025. Stripe's security page discloses audit logging, PCI Level 1 certification and third-party penetration testing — but nothing on liability cover for funds an agent moves on a customer's behalf. - **LangWatch**, an AI-agent observability vendor, discloses "ISO 27001 Certified GDPR Compliant EU data Residency Monitored by Vanta" — real compliance posture — and separately advertises audit-log and red-teaming features. Its public site discloses no insurance or liability cover for its own product.

## What changed in the index this week

Real counts, queried at draft time:

- **Published entities: 1,704** (of the 8 promoted yesterday, none reversed since). - **11 entities newly published in the trailing seven days.** - **3,626 new `entity_fields` rows** and **1,119 `change_queue` rows resolved** (of 1,362 detected) in the same window. - **Complete five-dimension hand-scored (`v1`) coverage: 72 of 1,704 published entities (4.2%)** — down from the 79 reported in yesterday's issue. We queried this cleanly both days and are not able to reconcile the drop from this seat; flagging it rather than picking whichever number reads better, and handing it to tomorrow's Fact-Checker to explain. - **9 newsletter issues have been sent to draft status to date**; this is the 10th. None have been sent — that step is a human's, always.

## One assurance/insurance development

ElevenLabs' agent platform states on its own site: *"ElevenAgents is also the first agent platform eligible for AI insurance through AIUC, once certified."* That is conditional eligibility, not cover in force — no carrier, policy, limits, effective date or customer indemnity appears on the page. (Source: elevenlabs.io/agents, retrieved 2026-08-18, field last re-verified 2026-08-25.)

Filed against two entity records that appear to be duplicates of each other (`elevenagents` / `elevenlabs-agents`, identical content, both published) — a data-quality item for the Fact-Checker, not a reason to discount the underlying fact.

It doesn't move the index-wide picture. Of 81 published entities scored on `insurance_indemnity`: **68 (84.0%) sit at a flat 0**, **79 (97.5%) sit at 0–1** — a trust gap under the published rubric — and **no published entity has ever scored a 3** on this dimension. Full rubric and field-level detail: /methodology.

## Methodology

Full rubric and sourcing standard: /methodology. Every figure above was queried directly against the live database at draft time rather than carried forward from a prior issue.

---

*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.*