Published report
OpenClaw 2026
Governed vs ungoverned AI agent behavior in a controlled 24-hour run, with stop behavior, approval mediation, and evidence quality measured side by side.
CAISI
Independent research and operating notes on AI agent governance.
CAISI Research
Reports, live builds, and artifact-backed findings
This is the canonical entry point for CAISI report pages. Use it when you want the measured artifact, the methodology, and the exact scope of a study before moving to interpretation or implementation guidance.
Published research
Published report
AI Tool and Agent Sprawl 2026
A public GitHub subset report focused on AI tool visibility, approval opacity, evidence posture, and governance readiness.
Archive and live-build context
Live build archive
AI Tool Sprawl Q1 2026
The earlier live-build page for the flagship sprawl study remains available as background on the study's build phase and publication posture.
Interpretation layer
The CAISI blog
The blog turns these reports into operating notes, role-specific lessons, and implementation guidance without changing the evidence base.
Start by role
AppSec
Start with runtime control evidence
OpenClaw is the fastest path if you want one measured example of where stop, approval, and proof break or hold.
CISO
Start with approval and evidence posture
The sprawl report is the better entry point if you need to reason about visible adoption, approval opacity, and proof quality across public artifacts.
Platform
Move from artifact to operating model
Read a report first, then move into the Operating Notes collection to see the underlying repo, workflow, and proof patterns.