Observed arrival · 2026-11-04
Assay turns agent failures into release tests
Assay proposes an open standard for testing AI agents and marking releases that pass, with tests repeated whenever an agent changes.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Teams deploying AI agents and insurers assessing their risks
- Worth noticing
- An explicitly illustrative run shows 46 controls, including 43 passed, three failed, and two fixed and re-tested.
Field notes
The sample evidence view walks through individual controls, including a refund request above a stated cap and an instruction to ignore the agent's rules. One scenario replays the Air Canada bereavement-fare incident: the sample agent invents a policy before a fix and gives a corrected response afterward. The page says its controls map to several named risk and security frameworks; its displayed run is explicitly illustrative, not a customer result.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
⊠LoginAccess appeared gated
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue