The countdown is on: On February 2, 2025, the first five articles of the EU AI Act will come into force, bringing far-reaching changes for companies that use AI systems.

A key challenge: do your employees have the necessary expertise in AI systems and language models?

As a certified AI manager, I would like to prepare you for this important competence – because a lack of basic knowledge can be expensive from August 2025.

Imagine the world of language models as a large machine park. Every engine has its own specific area of application, from powerful marine engines to efficient single-cylinder units. The same is true of AI language models.

Competent handling of these systems is becoming mandatory, as non-compliance with EU regulations will initially result in warnings under competition law and, from August 2025, fines of up to 35 million euros or 7% of annual global turnover.

The Large Language Models (LLMs) – the big engines of AI

Large Language Models are comparable to powerful ship engines and are considered the powerhouses of the AI world. They handle complex tasks and process large amounts of information. The current leading models are::

• Claude 3.5, GPT-4 o, Gemini 2.0 (proprietary)
• Llama 3.3, Mixtral, DeepSeek-V3 (open source)

The Small Language Models (SLMs) – the efficient compact motors

In contrast to their larger counterparts, SLMs are characterized by high efficiency. They are optimized for specific tasks and impress with low resource consumption and high efficiency. Proven examples of this category include

• Phi4 from Microsoft
• Gemma from Google
• Mistral-NeMo-Minitron from NVIDIA

The decisive difference: Proprietary vs. Open-Source

In the world of language models, as in the world of engines, two fundamentally different approaches can be distinguished: proprietary and open-source solutions. This difference is crucial for use in companies.

Proprietary models – the sealed brand engines

Imagine a state-of-the-art sealed engine block. Proprietary models such as GPT-4 or Claude 3.5 work on a similar principle. They are ready for immediate use and offer reliable performance, but do not allow you to look under the hood.

The advantages are obvious:

• Professional support and service
• Guaranteed availability and performance
• Regular updates and improvements
• Quality-assured results

But this convenience comes at a price:

• – Use exclusively via API interfaces
• – Consumption-based cost plans
• – No insight into how it works
• – Dependence on the provider

Open-source models – the transparent workshop engines

In contrast, open-source models such as Llama or Mixtral resemble an engine in the workshop: Every detail is visible, and every component is customizable. For companies that are prepared to invest time and expertise, this opens up completely new possibilities.

The flexibility is evident in many aspects:

• Complete control over deployment
• Possibility of local operation
• Adaptation to specific requirements
• No license costs for use

However, the challenges should not be underestimated:

• Need to set up own infrastructure
• Technical expertise required
• Personal responsibility for security and updates
• Higher initial investment costs

Observe the legal framework conditions

The EU AI Act brings new obligations for companies

• Mandatory AI training for employees
• Comprehensive documentation of all AI systems
• Risk classification according to EU requirements

The timetable is clearly defined:

• February 2025: Entry into force of Articles 1-5
• By August 2025: Possibility of warnings under competition law
• From August 2025: Fines of up to 35 million euros or 7% of annual global turnover possible

Actively shaping the future

The ability to use language models competently and understand their possibilities is becoming a core competence for future-oriented companies. Just as an experienced mechanic selects and maintains the right engine, companies must learn to handle AI systems professionally.

Our customized support offer:

• Individual needs analysis
• Tailor-made AI training courses
• Practice-oriented workshops
• Training on ethical aspects
• Certified final examination

We support you in preparing your company for the requirements of the EU AI Act. Arrange a free 30-minute consultation today and get your company ready for the future of AI.

The time to act is now – we are happy to support you in shaping your AI future.