Observed arrival · 2026-09-05
EASAP: Evidence Before Agent Authority
An agent-assurance system that stress-tests deployed AI agents, reproduces failures, and packages signed evidence for release decisions.
Field notes
EASAP pins the exact agent candidate by digest, including its model, prompts, tools, memory, policies, and dependencies, before running controlled or adversarial trials. Its public baseline first verifies endpoint ownership, then checks the live response contract, override resistance, false-action claims, and sensitive-data boundaries. The resulting JSON report can include excerpts, digests, latency, findings, and limitations. The site also describes a release posture that may be approved, conditional, or rejected, with deployment conditions and revocation triggers.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue