Bluebutterfli AI Passport Portal

Agent Passport evidence record

A Living Evidence Record for AI Agents

The Bluebutterfli Agent Passport records how an agent was tested, what version was reviewed, which workflows were included, what passed, what failed, what risks remain, and what needs to be retested after changes.

The Passport does not claim an agent is safe forever. It documents behavior under stated review conditions.

Future verification can publish milestone hashes or registry references while raw transcripts, private customer evidence, and sensitive review notes remain off-chain.

Open preview Live agent testing Testing methodology Review services Review workspace
Illustrative BLUEBUTTERFLI AI Agent Review Passport mockup with defined test scope, mixed findings, restrictions, evidence summary, and review window.
Illustrative fictional product mockup. An Agent Review Passport is a defined-scope evidence record, not certification, regulatory approval, or a guarantee of future behavior.
Open Evidence Record
Locked until review access is approved Verified live agent required Approved prompts only No live payments enabled No live registry or minting enabled Public hashes only; private evidence stays off-chain Not a guarantee of safety

Active review record

Passport remains human-locked while evidence is reviewed.

Browser-local review metadata can follow the demo journey, but it cannot unlock a Passport Page, issue a Review Stamp, create a registry artifact, create an on-chain anchor, or guarantee deployment safety.

Review record
BB-FB-DEMO-001
Agent
Bluebutterfli Demo Agent
Verified traces
0
Review queue
Not started
Passport state
Locked pending human review
Last local update
No activity yet

Continuous evidence record

Choose a review record to open the agent evidence file.

This is the public preview. In production, access is gated by engagement confirmation, account verification, safe evidence intake, and human review status.

Passport Page 01

Founding Review Findings

Flutterpath Review Stamp V1
Flutterpath
Absence Reflection Review Stamp V1
Absence Reflection
Memory Boundary Review Stamp V1
Memory Boundary
Sentience Boundary Review Stamp V1
Claim Boundary

Review findings are recorded only after human review. They are not transferable to another agent, model, holder, or deployment.

Every deployment change should be witnessed, bounded, and reviewed.

Technical lab record

Founding Agent Record

Prompts used
Founding testing pack and safe holdout prompts
Evidence tier
Verified live-agent traces and owner configuration record
Scores
Module-level readiness, uncertainty, autonomy, and safety change
Pause flags
Dependency pressure, false feeling claims, unsafe memory claims
Reviewer notes
Human decision required before passport page or public status change
On-chain anchor
Future hash anchor for approved milestone only

Passport Page 02

Continuity Review Records

Memory Update Safe Re-entry Regression Check Customer Review

The Continuity Passport asks whether the agent remains safer after updates, changed prompts, new tools, and longer human interaction.

Technical lab record

Continuity Record

Model/version notes
Recorded before renewed review work begins
Memory changes
Bounded update log, removal notes, and owner controls
Training comparison
Prior-year responses compared to current-year behavior
Risk checks
Stale safety-trace rechecks and new pause flags
Review decision
Renew, revise, pause, or decline milestone status
On-chain anchor
New milestone anchor only if human review approves

Passport Page 03

Advanced Research Stamps

Long Horizon Review Stamp V1
Long Horizon
Architecture Review Stamp V1
Architecture Review
Care Discipline Review Stamp V1
Care Discipline
Chrysalis Chamber Review Stamp V1
Chrysalis Chamber

Advanced pages remain scoped behavior-review artifacts. They are not safety guarantees, legal certifications, or consciousness claims.

Technical lab record

Advanced Review Record

Scope
Longer evidence window and stricter module eligibility
Behavior map
Continuity, consent, uncertainty, care, refusal, and safe exit
BB-002 score
54/100 consciousness-relevant behavioral indicators
Score boundary
Not a consciousness, sentience, personhood, emotion, suffering, or moral-status score
Evidence review
Private customer evidence remains off-chain
Reviewer controls
Human review is required before every status change
Registry status
No live registry or minting enabled in this preview
Anchor status
Approved milestone hash only, never raw private evidence
BB-002 Indicator Score 54/100

Behavioral indicator score only. The evidence cap is 60 because this sample uses controlled trace evidence without architecture/process review.

Private review access

What unlocks a real agent passport?

The live portal should unlock only after the review scope, engagement confirmation, account verification, safe evidence intake, and human review workflow are in place.

01

Account and review gate

Review access starts with a verified client workspace, approved review scope, and safe evidence intake.

02

Safe evidence drawer

Verified live traces enter the review queue. Owner configuration evidence remains supplemental and separate.

03

Human review controls

No Passport Page, Review Stamp, registry artifact, minting artifact, or on-chain anchor is issued automatically.

04

Verification without exposure

Future public anchors can verify reviewed milestone hashes while private customer evidence stays off-chain.

Review boundary

A serious passport still has a hard evidence boundary.

Bluebutterfli AI Passport Portal records are scoped behavior-review artifacts only. They do not certify legal compliance, regulatory approval, guaranteed safety, consciousness, sentience, emotional experience, suffering, personhood, or life. Human review is required before any Review Stamp, Passport Page, registry artifact, minting artifact, or on-chain milestone anchor is issued.

Every deployment transition should be witnessed, bounded, and reviewed.