01 / 03 · PURPOSE
What should Mikis help with?
Together, we describe the specific task: what information is used, who works with it and what decisions are involved?
Classification depends on the intended use, not just the server location.EU AI Act
Use AI responsibly in your organisation.
A clear purpose, appropriate access and an informed team form the foundation. We clarify the requirements for your specific task before implementation.
SECURE. CLEAR. PRACTICAL.
Set clear rules for AI from the start.
02 / 03 · FRAMEWORK
Agree requirements and responsibilities.
We document data flows, access permissions and intended outputs. The responsible advisers assess the legal classification, and we align the technical setup with it.
The RTR discussion concerned the Mikis concept presented at that meeting.03 / 03 · TEAM
Apply the rules in everyday work.
Your team learns about capabilities, limitations and data rules. Practical exercises cover checking sources and results, and knowing when expert review is needed.
Our training supports your organisation’s AI literacy measures.What high-risk means
High-risk is a legal classification based on purpose and area of use. Confidential data alone does not make an assistant high-risk. A human approval click does not automatically remove the classification. To get started, we select a clearly described task with you and record the data, people involved and intended decisions. This gives your adviser a basis for legal classification; we align the technical setup accordingly.
Applicant assessment
Selecting, filtering and evaluating job applications are expressly listed purposes.
Medicine
Emergency healthcare triage is expressly listed. Medical software may also fall under the rules for regulated products, subject to conditions including a required third-party assessment. Even for a summary of medical findings, the diagnostic or therapeutic intended purpose needs assessment.
Law and tax
Ordinary practice work is not automatically high-risk. The listed administration-of-justice purposes particularly concern assistance to judicial authorities and comparable alternative dispute resolution.
Transparency involves three different questions
Do people interact with AI?
They must be informed about direct interaction unless it is already obvious.
Is the origin technically detectable?
Within the scope of the obligation, generative system providers must ensure machine-readable marking and detectability. Exceptions and technical feasibility must be considered.
Will content be published?
Visible disclosure particularly concerns deepfakes and certain texts on matters of public interest. Such texts have exceptions for human review or editorial control with editorial responsibility.
Not every letter drafted with AI therefore needs an AI label. Conversely, a visible chat notice does not automatically fulfil the technical marking obligation. Together, we record which outputs your team creates and how they will be used. With the responsible parties, we identify which notices or technical measures should be configured within the agreed scope.
YOUR NEXT STEP
Discuss your deployment systematically.
Further information
Official information on the EU AI Act and guidance from the Austrian AI Service Desk.