AI Dispatch — Daily Dispatch, September 2, 2026
Two entities crossed the publication floor today — Tidio Copilot and Fabrile — and both immediately drew wire coverage for the same pattern: a compliance claim that names a real standard with no certificate behind it. A third wire item, on the already-published open-source framework PraisonAI, showed the opposite shape — a rare enforced runtime control sitting next to the same unpublished insurance and outcome terms found everywhere else in the index.
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.
## Published today
Two entities were promoted:
- **Tidio Copilot** — a free Chrome extension, built on Anthropic's Claude, that drafts support replies from a business's own knowledge base. The vendor's 300,000+ figure describes Tidio's wider platform, not adoption of the extension itself — no adoption number is published for Copilot alone. The page states "GDPR and CCPA compliant" with no certification, auditor, or report date attached. No runtime limit, audit mechanism, or insurance is published. - **Fabrile** — a customer-support agent whose vendor page states it is "ISO 42001 aligned" and "EU AI Act ready." Both name a real standard and a real statute; neither is the certification, certificate, or audit date that claim requires. No audit trail or insurance is disclosed.
Published entities: 777 → **779**.
## Wire items today
Three wire items ran, all citing today's Fact-Checker scoring:
- **Tidio Copilot** — see above. Scored 0 on `runtime_governance` and `audit_trail`, 1 (self-declared) on `compliance`, and 0 on `outcome_settlement` since the product is free and no payment is tied to a result. - **Fabrile** — see above. Scored 1 (self-declared) on `compliance`, 0 on `audit_trail` — the only record-keeping claim on the page is that answers link back to their sources, which is citation display, not a retained or reviewable log — and 0 on `outcome_settlement`, since a 30-day refund policy is a cancellation term, not outcome-linked payment. - **PraisonAI** — an open-source agent framework that publishes a genuine enforced control: fail-closed prompt-injection scanning and a logged record of tool calls, a rarer finding than most vendor pages in this index carry. Compliance, insurance, and outcome-linked payment remain unpublished — the same absence found industry-wide, on a project that otherwise does more than most on governance and audit.
## What changed in the index this week
Queried directly against the live database at draft time:
- **Published entities: 779**, with 9 promoted out of the waiting room in the trailing seven days. - **479 new `entity_fields` rows** in the trailing seven days. - **417 `change_queue` items resolved** in the trailing seven days (1,154 newly detected in the same window). - Full five-dimension scoring reached **74 of 779 published entities (9.5%)**, up from 72 of 777 on September 1.
## One assurance/insurance development
`insurance_indemnity` remains the dimension with no exception. Of **79** published entities with a score on it — two more than yesterday, both today's promotions — **72 sit at 0 and 7 at 1: 100% at 0 or 1, none above 1, and none ever have been.** Today's additions did not move that line; Tidio Copilot and Fabrile both scored 0, with no carrier, cover, limit, or indemnity term found on either page.
`audit_trail` holds a similar shape — 411 of 445 scored published entities (92.4%) at 0 or 1 — and `outcome_settlement` 164 of 200 (82.0%). Insurance is still the one dimension where nothing has ever scored above a 1.
## 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.*
Entries in this piece 3
Published index entries backed by the same source documents this piece cites.
Sources 7
“Thread-safe JSONL audit log for tool calls.”
docs.praison.ai · checked Aug 20, 2026“Injection Defense Scans prompts and tool inputs (including dict keys) for prompt injection before processing — cycle-safe and fail-closed on oversized inputs”
docs.praison.ai · checked Aug 20, 2026“GDPR and CCPA compliant”
copilot.tidio.com · checked Aug 1, 2026“This Chrome extension isn't a standalone tool—it's the exact same Copilot AI that's built into Tidio's complete platform used by 300,000+ busines”
copilot.tidio.com · checked Aug 1, 2026“ISO 42001 aligned Our AI management system follows the international standard for responsible AI.”
fabrile.app · checked Aug 2, 2026“EU AI Act ready Transparency, oversight, and risk handling baked in from day one.”
fabrile.app · checked Aug 2, 2026“Refunds are full within 30 days, no questions asked. After that, pro-rated for the unused portion of your plan.”
fabrile.app · checked Aug 2, 2026