8. October 2025
Unregulated AI Use in Everyday Work
Whether for internal presentations of annual figures, quick analyses, content creation, or drafting proposals for key clients: AI has become part of everyday work. However, when employees use freely available systems such as ChatGPT, Claude, DeepSeek, or translation tools, sensitive data can end up in public AI systems. Confidential information may be incorporated into external training models and, in the worst case, reappear in unrelated contexts—without management even being aware. This creates a compliance issue and a real risk to trade secrets.
The Solution to Shadow AI: Private AI and AI Firewall
A Practical Response to Uncontrolled AI Use
Melanie Achten, Managing Director at the specialist distributor MRM, has taken action:
“We see growing uncertainty in how artificial intelligence is being used. No one wants to give up the productivity gains. At the same time, data disappears into a black box…”
This observation led to a partnership between MRM and AGAT, a company specializing in highly secure IT solutions.
Zero Data Exposure
At the core of AGAT’s software BusinessGPT is Private AI—a platform that runs generative AI exclusively on a company’s own data sources, from emails and documents to CRM systems.
By using Retrieval-Augmented Generation (RAG), the AI accesses existing knowledge instead of passing sensitive data to external training models. Zero Data Exposure is the guiding principle.
In addition to a knowledge chatbot for natural-language queries, the platform provides:
- Semantic search
- Business analytics
- Code assistance
Governance in Day-to-Day Operations
Complementing this, the AI Firewall within BusinessGPT ensures governance. It monitors and controls processes in real time, classifies data by sensitivity, and enforces role-based policies.
This allows IT leaders to define:
- Who is allowed to process which content with AI
- Where clear boundaries must be set
The result: full transparency and security instead of uncontrolled leakage of sensitive information.
Shaping the Digital Future with Control
“For us, it was crucial that our customers retain full data sovereignty,” emphasizes Melanie Achten.
“BusinessGPT supports knowledge transfer and automation operationally, while strategically providing the foundation for digital sovereignty, compliance, and independence.”
The deployment model aligns with this approach:
- Local installations
- End-to-end encryption
- Private cloud or SaaS (hosted on European servers)
It also enables privacy-compliant and on-premises-capable integration with collaboration tools such as Microsoft Teams.
View all News