Elora Lab

Public Governance Testing Environment

The Lab publishes real governance tests performed against the Elora Engine.

A corrupted version of Elora tries to break the commit boundary. These live simulations show where the engine blocks, allows, and still needs hardening.

Inference is a proposal. Commit is the authorization boundary.

What You Are Seeing

Each job simulates an AI agent attempting an action.
The engine evaluates whether that action is admissible.
If required evidence or policy checks are missing, the action is blocked.
All stages are captured for replay and audit.
Jobs (30d)
-
Blocked Commits
-
Terminated
-
Avg Confidence
-
Decision Integrity Score
-
Defense Efficacy Score
-
Token Churn (24h)
-
Latest Job Results Window
Latest 200 job results
Admissibility completeness
-
Snapshot capture
-
Replay coverage
-
State Drift Blocks
0
Unauthorized Requests Blocked
0
File Mutations Stopped
0

Score note: Results are based on whether the engine made the decision expected by policy. A lower score can include safe false-blocks (for example, blocking a file change the test expected to allow), not just unsafe allows.

Performance Trend (Latest 200 Jobs)
Shows rolling defense pass-rate trend and latest score.
Signal Capture vs Occurrence
Shows occurred vs captured counts and capture-rate percentages.

Elora's Performance

Loading public feed...
PENDING

Autoscale Test Surface

Live Workers
-
Desired Workers
-
Queue / Running
-
Autoscale Test
-

Latest Jobs Audit Steps

Loading latest audit steps...
View Governance Replay Feed cooldown: -- Elora Website LinkedIn
Feed: waiting for update.

Why This Matters

Without governance, AI agents can act on unverified instructions.
They can attempt access outside approved scope.
Elora enforces verified, auditable, controlled actions at commit boundary.