Clinical safety

VibeCheck.luxury is built so the AI drafts and a person always decides.

Where the software stops, who it escalates to, and the four points in the workflow a human has to pass.

Quick answer: VibeCheck.luxury keeps clinical authority with the licensed clinician by design. Every AI-generated clinical output is reviewed and signed before it enters the record, four gates in the workflow require a person, and the AI can never clear a risk flag. Crisis resources are always on in the client app and a flag escalates to a human — but this is not a crisis service.

Data security How the loop runs →

THE FOUR GATES

Where a human has to decide

Every AI-generated clinical output is reviewed and signed by the licensed clinician before it enters the record.

① Note attest

The therapist signs every note. The AI can never finalize one.

② Claim attest

No claim submits without a human.

③ Treatment plan

The clinician's, with co-sign supported.

④ Risk acknowledgement

A flag escalates to a person. The AI can never clear it.

These are gates, not preferences. There is no setting that removes them, which is the difference between software that drafts and software that decides.

THE COMPANION

What it will not do

The companion your client talks to is channeled through rules you write — what it can discuss, how it responds to urgency, what comes back to you. It does not present itself as the clinician and it does not carry the clinical relationship. An AI therapist pretends the relationship belongs to the model; this is the opposite design.

It does not diagnose. It does not change a treatment plan. It does not decide that someone is safe. Those are clinical judgments and they stay with the person holding the licence. How supervision actually works →

CRISIS

The floor, and its limits

Crisis resources are always reachable inside the client app, and a risk flag routes to a person rather than to a model. That is a floor under the product, not a substitute for care.

VibeCheck.luxury is not a crisis service. If someone is in danger, the path is 988 (Suicide & Crisis Lifeline) or the nearest emergency department. We say that on the client-facing page too, in the same words, rather than burying it.

THE LIMITS

Said plainly

Between-session signal is not continuous monitoring, and nobody is watching a screen. A client can stop using the app, and a gap in check-ins is information rather than an alarm. The pre-session brief summarises a week; it does not replace your judgment about what that week meant.

We do not make outcome claims. Whether any of this improves care in a given case is a clinical question about a specific person, and it stays yours to answer.

QUESTIONS

Questions clinicians ask

Is the AI companion a therapist?

No, and it is built not to become one. It runs on rules the clinician writes, it does not position itself as the clinical relationship, and it cannot make a clinical decision. The therapist is the clinician; the software is a channel.

What happens if a client says something urgent?

Crisis resources are always reachable inside the client app, and a risk flag escalates to a person. The AI can never clear a risk flag — that is one of the four gates it cannot pass. This is not a crisis service: 988 and emergency services are the crisis path.

Can the AI finalize a note or submit a claim?

No. The therapist attests every note and the AI can never finalize one; no claim submits without a human; the treatment plan is the clinician's with co-sign supported; and a risk flag escalates to a person. Those are gates, not settings — there is no configuration that turns them off.

Does the client know what their therapist sees?

They should, and the page written for them says so plainly and tells them to ask. Consent and boundaries are the clinician's to set, and the client-facing page encourages the conversation rather than assuming it happened.