What it means for your organization
Your obligations under the EU AI Act are determined by two things: the risk class of your assistant (what it is used for) and your role (whether you share it outside your organization).
Below are the three most common scenarios for organizations using Intric, along with the full checklist of requirements for each.
Not sure which scenario applies to you?
Scenario A: Deployer, not high-risk
Section titled “Scenario A: Deployer, not high-risk”This is the most common scenario. Your assistant is used internally for tasks that do not fall in a high-risk area, or where a valid exception applies.
Examples: Drafting job postings based on templates, summarizing reports, answering questions about internal rules and procedures, quality-checking documents after a human has already completed the assessment.
Your obligations:
- AI literacy: Ensure that all users have sufficient knowledge to use the platform safely and make informed decisions about the content Intric generates.
Scenario B: Deployer, high-risk
Section titled “Scenario B: Deployer, high-risk”Your assistant operates in a high-risk area, such as access to welfare services, employment and recruitment, or education, and it influences decisions that affect individuals.
Examples: An assistant that supports case management for social welfare applications, or that sorts and evaluates candidates in a recruitment process.
Your obligations:
- Human oversight: Ensure real human oversight by people with the right authority and competence. A human must always have the final say.
- Notify employees and unions: Inform employee representatives and employees before the system is put into use in the workplace.
- Notify affected persons: Persons affected by AI-assisted decisions (e.g. applicants) must be informed that AI is being used as decision support or in the decision-making process.
- Correct input data: Ensure that only relevant and appropriate data is used as input to the assistant.
- Use in accordance with instructions: The assistant must be used in accordance with Intric’s instructions for use.
- Monitoring and logging: Monitor use systematically and log relevant events.
- Register with authorities: Register the system in the relevant public register (e.g. Nkom in Norway).
- DPIA: Carry out a Data Protection Impact Assessment if the assistant processes personal data.
Scenario C: Downstream provider, high-risk
Section titled “Scenario C: Downstream provider, high-risk”You have built a high-risk assistant and share actual system access with another organization, for example via Intric’s widget functionality. You then take on provider-level obligations for that assistant.
Note: Sharing a prompt via the Assistant Library does not make you a downstream provider. It only applies when you give another organization direct access to an assistant you have built.
Example: Your organization has built an assistant for assessing building permit applications and gives a partner organization access to it through an inter-organizational collaboration.
Your obligations include everything in Scenario B, plus:
- Quality management: Establish and maintain a formal quality management system for the assistant.
- Risk assessments: Carry out and document formal risk assessments before and during sharing.
- Data governance: Establish routines for data management and data quality.
- Registration: Register the assistant as a high-risk system with the relevant authority.
- Monitoring and logging: Implement systematic monitoring of how the assistant performs and is used by the receiving organization.
Transparency obligations
Section titled “Transparency obligations”Even if your assistant is not high-risk, you may have transparency obligations under AI Act art. 50. Check whether any of the following apply:
| Situation | Obligation |
|---|---|
| The assistant interacts directly with people (e.g. via widget, chatbot, or citizen portal) | Users must be clearly informed they are interacting with an AI system |
| The assistant generates text intended to inform the public on matters of public interest | The text must be labeled as AI-generated |
| The assistant recognizes emotions or performs biometric categorization | Affected persons must be informed |
| The assistant generates or manipulates images, audio, or video that could constitute a deepfake | The content must be labeled as AI-generated or manipulated |
Want a guided assessment?
Section titled “Want a guided assessment?”The AI Act Screening Assistant in the Assistant Library walks you through the full decision tree step by step, in Norwegian or English. It produces a documented conclusion with legal references that you can save and use as compliance documentation.