· Compliance
The AI Act hits Norwegian SMBs in August 2026.
You don't need a legal office in Brussels. You need an assessment, a risk classification, and documentation that holds up to scrutiny.
What the law says, briefly
The EU AI Act entered into force on 1 August 2024. The most important parts for most businesses are enforced from August 2026.
- · AI systems are classified into risk categories: prohibited, high-risk, limited risk, minimal risk.
- · High-risk systems (recruitment, credit scoring, health) require a DPIA-like assessment, documentation, human oversight, and logging.
- · Generative AI systems (ChatGPT, Claude, Gemini) need explicit transparency toward users when output can be mistaken for human-made.
- · Breaches carry fines up to EUR 35 million or 7% of global turnover, whichever is higher.
What sets Tenki apart
We aren't a law firm. We're a technical team that has implemented local AI for Norwegian SMBs and read the AI Act text with practical implementation in mind, not academia. The difference in practice:
Traditional advisory
- · Workshops and PowerPoint
- · 80-page report that ends up in a drawer
- · Focus on paper, not code
- · 200,000+ NOK
Tenki
- · Interviews + AI inventory
- · Risk classification per system
- · Implementable documentation
- · Shorter engagement, lower threshold
Our own practice
If we recommend that a customer does it, we have to be able to do it ourselves first. Tenki publishes what Hugin and Munin do, logs interactions for quality assurance with a clear retention policy, and builds AI Act practice into how we develop new AI features.
Transparency
Hugin (local LLM) and Munin (cloud LLM) are clearly labelled at tenki.no/chat with what processes the data.
Logging and retention
Chat interactions are stored for 90 days for quality assurance, described in the privacy policy.
How we work with you
- 1. Assessment. We interview key people and build an inventory of every place you use AI. This includes hidden cases like "we paste customer replies into ChatGPT" or "our supplier uses AI behind the scenes."
- 2. Classification. Each AI system gets a risk classification under the AI Act, linked to specific legal references. We point out which require immediate action and which are safe today.
- 3. Documentation. We deliver templates and fill in DPIAs, transparency declarations, and a register that matches AI Act requirements. Not generic law, but specific to your systems.
- 4. Follow-up. A quarterly check-in where we review changes, new tools, and that the documentation still holds. Included in the package for the first year.
When should you start?
If you use generative AI in customer contact or internal operations: start now. Classification takes 2–4 weeks, documentation another 2–3 weeks. That gives you plenty of time before enforcement begins in August.
If you're in doubt about whether you actually use AI that falls under the law, take the maturity test — five minutes, gives you a starting point for the conversation.