The AI Act came into force in August 2024, but many companies are still unaware that the new regulations directly affect them.
If you sell SaaS, implement AI in customer processes or build decision automation tools, The AI Act applies to your business.
Below you will find a practical explanation of what this means in B2B business – without legal jargon, but with concrete obligations.
1. What is the AI Act and why does it apply to B2B companies?
The AI Act is the first comprehensive regulation in the EU regulating artificial intelligence.
It regulates not companies, but AI systems and their applications.
Why does it apply to B2B?
Because the regulations do not look at the size of the company, but only at:
- what the AI system does
- what effect does it have on people
- in what area is it used
Examples of systems covered by the AI Act:
- recruitment systems with AI
- customer scoring and risk assessment
- customer service chatbots
- recommendation systems in SaaS
2. Who is who in the AI Act?
The AI Act defines several roles that may overlap.
2.1 Provider (AI provider)
This is a company that:
- creates an AI system
- introduces it to the market
- sells it under its own brand
Most common: software house, SaaS, AI startup
2.2 Deployer
This is a company that:
- uses an AI system in its operations
- I don't sell it any more
- uses it operationally
e.g. e-commerce using AI to serve customers
2.3 Integrator (practical role)
Integrator:
- combines various AI systems
- adapts solutions to the client
This is not a formal category in the AI Act, but it plays an important business role.
3. Obligations of AI providers
If you are an AI provider, the AI Act imposes specific obligations.
3.1 Risk assessment and documentation
You have to:
- identify system risks
- describe their impact
- document countermeasures
3.2 Quality management system
The following processes are required:
- AI testing
- performance monitoring
- model updates
- responding to errors
3.3 Transparency
The user must know:
- that it uses AI
- how the system works
- what are its limitations
3.4 Human oversight
The system must enable:
- human interference
- suspending AI decisions
- verification of results
3.5 AI Systems Registry
For high-risk systems:
- obligation to report to the EU database
4. Transparency in the AI Act
Transparency is one of the key responsibilities.
What does the user need to know?
- that it uses AI
- what decisions does the system make
- on what basis does it work
Examples:
- the chatbot must inform that it is AI
- the recruitment system must reveal automation
- Customer scoring must be explainable
5. AI Risk Categories
The AI Act divides systems into 4 risk levels:
5.1 Unacceptable systems
Banned in the EU:
- social scoring of citizens
- behavior manipulation
- child abuse
5.2 High-risk systems
They concern, among others:
- HR and recruitment
- financial scoring
- critical infrastructure
- decision-making systems
5.3 Risk-limited systems
Transparency is mainly required:
- chatbots
- deepfake
- AI content generation
5.4 Minimum Risk Systems
No additional obligations:
- spam filters
- content recommendations
6. Human-in-the-loop and AI ethics
Human-in-the-loop means that:
humans must be able to control AI decisions
Examples:
- recruitment: AI analyzes CVs, humans decide
- scoring: a human can verify the result
Why is this important?
- reduces the risk of errors
- limits liability
- increases legal security
7. Common company mistakes (red flags)
1. No AI risk assessment
Companies are implementing AI without regulatory analysis.
2. Lack of provisions in contracts
Is lacking:
- responsibility
- AI usage rules
- compliance with the AI Act
3. No human review
The system works automatically in 100%.
4. Lack of legal documentation
There is only technical documentation, no compliance.
5. Lack of information for users
The user does not know that he is using AI.
8. When should I check compliance with the AI Act?
It's worth doing this if:
- you implement AI in HR, scoring or data analysis
- you use AI in SaaS
- you automate customer processes
- you sign contracts with AI suppliers
- you are developing a technological product
Summary
The AI Act changes the way tech and B2B companies use AI.
The most important duties are:
- risk assessment
- transparency
- human supervision
- documentation
- contractual liability
Companies that implement the AI Act correctly benefit from:
- greater legal security
- advantage in tenders
- customer trust
Want to check your AI's compliance with the AI Act?
If you want to verify your AI solutions or SaaS contracts against new regulations, it is worth conducting an analysis before risks arise.