AI Transparency and Responsible Use

You are talking to an AI

When you practice in Kulissa, or when you speak with Maya on the public site, you are interacting with an artificial intelligence system. The voice may sound human. It is not a human. It can interrupt, object, change tone and make mistakes.

We disclose this because it is honest, and because transparency duties such as Article 50 of the EU AI Act require that people know when they are interacting with AI in cases like ours. Product surfaces identify the buyer as AI. Maya is labelled as Kulissa's gatekeeper and as an AI buyer.

What the AI is for

Kulissa's AI buyers exist so that people can rehearse high-stakes conversations before a real customer, colleague or manager is on the line. Scores and feedback exist so that coaches and learners can see what happened in the transcript and improve.

What the AI is not for

• Not an employment, promotion, dismissal, credit, insurance or housing decision system
• Not a silent recorder of live customer calls (that is a different product category)
• Not a guarantee that a learner is ready; readiness is a human judgement informed by evidence

Customer organisations must not use practice scores as the sole basis for a decision that produces legal or similarly significant effects about a person. Human oversight stays with the Customer's managers. See the Terms.

How it works, briefly

1. Customer material (playbooks, documents, rubrics) is used to build scenarios.
2. A learner speaks; audio is transcribed; a model produces the next buyer turn and, after the session, a structured score validated against a schema.
3. Evidence for a score is checked against the transcript we actually have. A model claim about data we never supplied is rejected.

Data and training

Customer Content and Maya demo content are sent to model providers only to produce the response or score for that request. We do not use that content to train our own models, and we require providers not to train on it. Details: Privacy, Subprocessors.

Limitations

AI buyers can hallucinate product facts, miss nuance, or produce uneven difficulty. Rubrics and Customer review of generated scenarios are the control. Learners should treat odd output as a practice moment, not as ground truth about a real account.

Content moderation and safety

We prohibit uses listed in the Acceptable Use Policy. We may interrupt or refuse sessions that attempt to elicit illegal content or to abuse the demo.

Maya

Maya is a public demonstration and lead magnet. She is a hard buyer on purpose. Credits she awards are promotional, not a plan. Microphone permission is yours to grant or deny.

Changes

We will update this notice as the product and the law develop.

Contact

hello@kulissa.com with subject [Legal] AI. Operator: STEALTH, Paris and Zurich.

FAQ

Does a high score mean someone should be promoted?

No. Scores are coaching evidence. Humans decide.

Can the AI pretend to be a specific real person?

Scenarios may use fictional personas or role titles from Customer material. Impersonating a real private individual to deceive is prohibited under Acceptable Use.