EU AI Act: what B2B companies need to do now
The EU AI Act is the first comprehensive AI legislation in the world. AI literacy has applied since February 2025. In July 2026, the law was amended by the Digital Omnibus. The high-risk rules have been postponed to December 2027. In this article I explain what the law entails, what it means for your business, and what you need to do now.
AI & automation consultant. Helps B2B companies with lead generation, workflow automation, and AI training.
Why this matters now
The EU AI Act is not a future plan. It is law. Parts of it are already in effect. The AI literacy duty has applied since February 2, 2025 for every company that uses AI. Prohibited AI practices should have been stopped on that same date.
Since August 2, 2026, supervisory authorities oversee the rules, including AI literacy. The high-risk rules have been postponed. For standalone systems, such as AI in hiring, they apply from December 2, 2027. For AI in products already covered by other EU rules, they apply from August 2, 2028.
Meanwhile, many SMEs think the law doesn't apply to them. That is a misconception.
What is the EU AI Act in plain English?
The EU AI Act is a European law that regulates how AI systems may be developed and used. The goal is to ensure that AI is deployed safely, transparently and fairly, without unnecessarily hindering innovation.
The core of the law is a risk-based approach. AI systems are categorized into four levels based on how great the risk is to society. The higher the risk, the stricter the rules.
For most B2B companies that use AI for sales, marketing and operations, the lowest risk level applies. But there are general obligations that apply to everyone, regardless of risk level.
What changed with the Digital Omnibus?
In 2026, the EU amended the AI Act. It did so with the Digital Omnibus: Regulation (EU) 2026/1744. It has applied since July 27, 2026. The goal is simpler rules. More background is in the European Commission's questions and answers.
What changed:
What stayed the same:
The 4 risk levels explained
The EU AI Act categorizes AI systems by risk. This determines which rules apply to you.
Prohibited
AI that violates fundamental rights. Completely banned since February 2025.
Examples:
- Social scoring by governments
- Real-time facial recognition in public spaces (with exceptions)
- AI that recognizes emotions in the workplace or education
- AI that manipulates vulnerable groups
What does this mean for you? Not relevant for most B2B companies. But if you use AI to monitor or score employees, be careful.
High risk
AI that influences important decisions about people. Strict rules, postponed to December 2, 2027. For AI in products already covered by other EU rules: August 2, 2028.
Examples:
- AI in hiring processes (CV screening, ranking of candidates)
- Credit scoring and financial assessments
- AI in education (student assessment)
- Medical AI systems
What does this mean for you? Relevant if you use AI for HR processes, employee evaluation or financial decisions. Not for standard sales and marketing.
Limited risk
AI systems that require transparency. You must inform users that they're dealing with AI.
Examples:
- Chatbots on your website
- AI-generated content (text, images, video)
- Deepfakes and synthetic media
- Emotion recognition (where permitted)
What does this mean for you? Relevant if you have an AI chatbot on your website or publish AI-generated content. Make sure users know they're communicating with AI.
Minimal / Low risk
Most AI applications in B2B sales and marketing. No specific obligations, but best practices are recommended.
Examples:
- AI for lead generation and lead scoring
- Personalized email outreach
- Content creation with Claude or GPT
- Workflow automation with n8n
- CRM automation and data enrichment
What does this mean for you? This is where most B2B companies sit. No strict rules, but AI literacy is mandatory and documentation is recommended.
What applies to B2B sales and marketing tools?
If you use AI agents for sales , ChatGPT for emails, Claude for content, or n8n for workflow automation, you fall under the minimal or low risk category.
That means: no conformity assessment, no certification, no mandatory audit. But it doesn't mean you don't need to do anything.
What is mandatory:
What is strongly recommended:
AI literacy: how to arrange it in four steps
AI literacy means: your people understand what AI does. They know where it can go wrong. And they know when to think for themselves.
You don't need to arrange a course, exam or certificate. You decide how to approach it. These four steps get you started.
- 1
Map which AI tools your team uses, and what for.
Make a list. Which tools does your team use? And for what work? Don't forget the tools people picked themselves.
- 2
Decide per role what someone needs to know.
Someone who uses ChatGPT for an email needs to know different things than someone who puts customer data into an AI tool. Look at the tools and the risks per role.
- 3
Give a short explanation, set AI ground rules and appoint someone to keep track.
A short explanation per role is enough to start. Write down what is and isn't allowed. For example: which data must never go into an AI tool? And pick one person who keeps track.
- 4
Look back after a while and briefly record what you did.
Plan a moment, for example after six months. Does it work? Are there new tools? Briefly write down what you did. Then you can show it if someone asks.
Free AI literacy starter kit
Don't want to figure this out yourself? I made a starter kit. It contains:
- A checklist to work through step by step
- An overview per role: what does someone need to know?
- AI ground rules to fill in yourself
Want the starter kit? Send me a DM on LinkedIn with "AI-KIT". I'll send it to you. Open Nico's LinkedIn profile
Checklist: 10 things to arrange now
Start at the top and work your way down. Items with an orange marker are the most urgent.
1. Create an AI register
UrgentMap all AI systems your organization uses. Per system: name, vendor, purpose, risk level, responsible person and start date.
2. Take care of AI literacy for your team
UrgentMake sure everyone who works with AI understands what it does and where it can go wrong. This has applied since February 2025. No certificate needed. Do briefly write down what you did.
3. Draft an AI policy
UrgentAn internal document that describes how your organization uses AI, which rules apply, who is responsible and how you handle complaints.
4. Appoint an AI responsible person
Someone who owns the AI policy, maintains the register and is the point of contact for questions. Doesn't need to be a full-time role.
5. Check your AI vendors
UrgentAsk your vendors (OpenAI, Anthropic, etc.) about their compliance. How do they process data? Where are the servers? Do they have a data processing agreement?
6. Check your data processing agreements
UrgentEnsure you have a data processing agreement with every AI vendor that complies with GDPR and aligns with the AI Act. Pay attention to data storage and model training.
7. Be transparent to users
If you use AI chatbots or publish AI-generated content, inform your users. Add a disclaimer to chatbots and label AI content where relevant.
8. Keep a human in the loop
For decisions that affect people (such as lead scoring that determines who gets contacted and who doesn't), ensure a human makes or can override the final decision.
9. Document your AI usage
UrgentRecord per AI application: what it does, what data it uses, how the output is monitored and what you do when it goes wrong. This is your evidence during an audit.
10. Schedule an annual review
AI legislation evolves. Schedule at least one review per year of your AI register, policy and vendors. Adjust where needed.
Common mistakes
Thinking the AI Act doesn't apply to you
The most common mistake. 'We only use ChatGPT for emails, that doesn't count.' Yes, it does count. Any use of AI in your business operations falls under the law. The risk level is low, but the AI literacy duty applies to you too.
Not having an AI policy
Without a policy, there is no framework. Employees use AI tools at their own discretion, there's no overview of what's happening, and during an audit you can't show anything. A policy doesn't need to be long, but it needs to exist.
Ignoring AI literacy
This has applied since February 2025 and is the most often forgotten. The duty became lighter, but it hasn't gone away. A course or certificate isn't required. You do have to act on it. Briefly write down what you did.
Forgetting data processing agreements
If you use ChatGPT or Claude for business data, you're processing data via a third party. Without a data processing agreement, you're not compliant with GDPR, and neither with the AI Act.
Thinking the delay means you don't have to do anything
The high-risk rules have been postponed. The rest hasn't. AI literacy has applied since February 2025. Supervisory authorities have been watching since August 2026. And do you use AI for HR or credit scoring? Then you have until December 2027. That sounds like a lot, but such a project takes months.
What does it cost to become compliant?
Policy + Implementation
from €5,000Drafting an AI policy, creating an AI register, AI literacy workshop for your team, checking data processing agreements and setting up documentation structure.
AI Act Quick Scan (first step)
€750Inventory of your current AI usage, risk analysis per application, and a priority list with concrete actions. You'll know exactly where you stand.
Perspective: for most rules, the maximum fine is 15 million euros or 3% of your annual turnover. For prohibited AI, it is 35 million euros or 7%. A Quick Scan of 750 euros is a small investment to know where you stand.
How WaiBase helps with this
At WaiBase, I combine AI expertise with practical compliance knowledge. I don't just build AI automations, I also make sure they're compliant.
Also check out my AI consulting service and AI training for more information.
Timeline: what becomes mandatory when
EU AI Act enters into force
Prohibited AI practices must be stopped. The AI literacy duty applies.
Rules for general-purpose AI models (GPT, Claude, Gemini) apply
Digital Omnibus applies. AI literacy becomes lighter, the high-risk rules are postponed.
Most of the law applies. Supervisory authorities oversee the rules, including AI literacy.
Rules for standalone high-risk AI systems apply, such as AI in hiring or credit scoring.
Rules for high-risk AI in products already covered by other EU rules, such as machinery and medical devices.
Frequently asked questions about the EU AI Act
Does the EU AI Act apply to my SME?
Yes. The EU AI Act applies to every company that develops, deploys or imports AI systems within the EU. It doesn't matter how big you are. If you use ChatGPT, Claude or other AI tools in your business operations, you must comply with the rules that correspond to your application's risk level.
What is AI literacy?
AI literacy means your people understand what AI does. They know what it can't do and where it can go wrong. It covers everyone in your company who works with AI. How much someone needs to know depends on their role and the tools they use.
Is AI literacy mandatory?
Yes. Since February 2, 2025, companies must do something about the AI literacy of their people. The Digital Omnibus made that duty lighter. You must support the AI knowledge of your people. You don't have to guarantee a specific level per employee. Since August 2, 2026, supervisory authorities oversee this.
Do I need a certificate?
No. The law doesn't require a certificate, an exam or a mandatory course. You decide how you help your people. A smart move: briefly write down what you did. Then you can show it if someone asks.
What changed with the Digital Omnibus?
The Digital Omnibus (Regulation (EU) 2026/1744) has applied since July 27, 2026 and amends the AI Act. Two things changed. AI literacy became lighter: you support the knowledge of your people, but you don't have to guarantee a level per employee. And the high-risk rules were postponed: to December 2, 2027 for standalone systems, and to August 2, 2028 for AI in products already covered by other EU rules. The prohibitions, the four risk levels and the rules for chatbots stayed in place.
Are AI sales tools high risk?
Generally not. AI tools for lead generation, email outreach, content creation and CRM automation fall under low risk or minimal risk. They only become high risk when used for evaluating people in selection processes, credit scoring or employee evaluations.
What is an AI register and do I need one?
An AI register is an overview of all AI systems used in your company: which tools, what for, who is responsible, and which risk level applies. It's not mandatory for everyone, but it's a best practice recommended by the Dutch Data Protection Authority.
What are the fines for non-compliance?
The maximum fines are substantial. For deploying prohibited AI, it is 35 million euros or 7% of global annual turnover. For most other rules, the maximum is 15 million euros or 3%. In practice, enforcement for SMEs will start with warnings and improvement plans, but the risk is real.
When do which rules apply?
The law applies in steps. Since February 2, 2025, the prohibitions and the AI literacy duty apply. Since August 2, 2025, the rules for general-purpose AI models such as GPT and Claude apply. Since August 2, 2026, most of the law applies and supervisory authorities oversee it. The high-risk rules apply from December 2, 2027 for standalone systems, and from August 2, 2028 for AI in products already covered by other EU rules.
How long does it take to become compliant?
The basics (AI register, policy, training) can be arranged in 2 to 4 weeks. A full compliance process including vendor checks and documentation takes 1 to 3 months. The sooner you start, the less pressure as the deadline approaches.
What does it cost to become compliant?
A complete policy and implementation process starts from €5,000, depending on the number of AI systems and the complexity of your organization. Want to know where you stand first? Then start with an AI Act Quick Scan of €750. Compare that to the potential fines and reputational damage.
Do I need a Data Protection Officer (DPO)?
The EU AI Act doesn't require a specific AI Officer, but it's wise to appoint someone responsible for AI governance. If you already have a DPO for GDPR, that role can be expanded. For larger organizations, a separate AI responsible person is recommended.
Where can I find the official text of the EU AI Act?
The full text is available on EUR-Lex, the official legal database of the EU. The Dutch Data Protection Authority (AP) also publishes guidelines and practical guidance. For a readable summary, you can use this article as a starting point.
Do you know where you stand?
Book an AI Act Quick Scan and know within a week exactly what you need to do. No legal jargon, just a concrete action plan.
Book a strategy call