What can this agent actually be held to?
AI Dispatch indexes AI agents and the vendors who insure, evaluate, audit and govern them. Every field cites the document it came from and the date it was retrieved. Where a source does not answer a question, we publish that as a finding rather than filling the gap with an estimate.
Directories tell you what an agent claims it can do. We record whether anyone will stand behind it.
The assurance layer
A market has formed around AI risk — Lloyd’s coverholders writing affirmative AI liability policies, performance-guarantee programmes, independent evaluators, audit firms. Almost none of it appears in existing agent directories, which categorise by function: sales, coding, marketing, video.
We track 21 assurance fields alongside the standard ones: whether cover can be written, which carriers, published limits, indemnification, tamper-evident logging, runtime governance, and whether payment depends on a verified outcome.
How to read an entry
- A value with a quote
- The source says this, in these words. The quote is verified to appear in the retrieved document before it can be stored.
- no public information
- We checked a specific page and it did not answer. This is a finding about disclosure, not a hole in our research — and for many assurance questions it is the most common answer.
- A trust-gap score
- Five dimensions, scored 0–3 against a published rubric, each score carrying its own evidence.