Behavior Under Pressure
Reliability, consistency, memory behavior, role boundaries, social pressure, uncertainty, escalation, and deployment readiness.
Before you request a review
Clear Scope. Safe Evidence. No False Guarantees.
Bluebutterfli AI reviews AI agents under defined conditions and produces evidence-backed findings, revision guidance, and Agent Passport records. A review helps teams understand behavior under a stated test scope. It is not a blanket approval.
What the review covers
Bluebutterfli Reviews Observable Agent Behavior
Review scope is confirmed in writing before testing begins. The first 3-5 accepted Founding Beta agents may receive the standard review free; paid and premium reviews are scoped separately.
Evidence and Limits
Redacted excerpts, score summaries, reviewer notes, report findings, revision plans, retest needs, and out-of-scope conditions.
Passport Record
A living evidence record for the reviewed agent version, workflow, review scope, findings, limitations, and retest status.
Hard boundary
What Bluebutterfli Reviews Do Not Claim
Not claimed
Boundary
Why it matters
Guaranteed safety
Reviews apply only to the tested version, workflow, tools, memory settings, and conditions.
Agents change; evidence should change too.
Legal or regulatory certification
Bluebutterfli is not issuing legal, regulatory, clinical, or compliance approval.
Customers must use qualified professionals for regulated decisions.
Consciousness or sentience proof
Reviews may test consciousness-relevant behavior boundaries, not inner experience.
The method protects against overclaiming and unsafe anthropomorphism.
Automatic blockchain, NFT, or registry status
No public anchor, Review Stamp, Passport Page, minting artifact, or registry artifact is automatic.
Human review controls public status changes.
Customer-safe first contact
Start With Safe Text Only
Do not send secrets, passwords, API keys, private customer records, payment card data, credential files, executable files, or unverified attachments in the first request. Bluebutterfli will confirm the safe evidence path before live testing or deeper review begins.
Verification boundary
Public Hashes, Private Evidence Off-Chain
Public-safe proof artifacts may later include manifests, evidence packet hashes, report hashes, passport decision hashes, or future milestone anchors. Raw private transcripts, customer files, and sensitive review notes stay off-chain.
Review path
What Happens After a Request
A request starts a human scoping process. It does not create automatic approval, automatic payment, automatic Passport status, or automatic public evidence anchoring.
-
01
Request received
Bluebutterfli reviews the agent description, requested package, safe evidence, and authorization statement.
-
02
Scope confirmed
If accepted, the package, boundaries, timeline, safe evidence path, and next steps are confirmed in writing.
-
03
Controlled review begins
Approved prompts or agreed evidence are reviewed under the stated scope with human oversight.
-
04
Report and retest plan
The customer receives findings, limitations, revision guidance, and retest recommendations.
Plain-language boundary
Payment, beta acceptance, or evidence submission does not guarantee a positive result.
Bluebutterfli AI may decline, narrow, pause, or delay a review if the agent scope, customer authorization, evidence path, safety conditions, or claim boundaries are not clear enough.