AI transparency

What Kai is

Kai is a personal AI agent for each person at work. It runs inside the customer's company, on open models running locally. It helps with email, meetings and documents, then learns, under that person's guidance, how they handle routine tasks and takes on the busywork. The hard problems stay with people.

People know it is AI

People Kai interacts with are told they are dealing with AI, and content Kai writes can be marked as AI-generated, as Article 50 of the EU AI Act requires.

What Kai never does

• Read or infer emotions, stress, mood or health.

• Rate, score or rank people.

• Read screens or keystrokes. It works only from company work tools.

• Make decisions about anyone's job.

Why Kai is not a high-risk system

The EU AI Act treats some workplace uses of AI as high-risk: recruitment, decisions about promotion or dismissal, assigning work by people's behaviour or traits, and monitoring or evaluating performance. Kai is not offered for any of them, and our contracts and Acceptable use policy forbid them. Within one person's work, Kai does narrow, preparatory tasks that the person approves, and it does not profile people.

Acceptable use

People approve the risky actions

Low-risk actions can run on their own. Payments, client emails and access changes are prepared and wait for a person's approval, and every action is logged where the person it worked for can see it.

Our team

Everyone at Xamar AI who builds or supports Kai has had AI literacy training, as Article 4 of the EU AI Act requires.