Your employees are already using AI, often with private accounts and without knowing where customer data, prices or source code end up. Banning it rarely helps. A secure route is better: a central AI gateway with a privacy filter detects personal and confidential data, replaces it with placeholders before it is sent to a cloud model and restores it in the response.
What you gain
No sensitive data in the cloud
Personal and confidential content is automatically detected and replaced before it is sent.
Controlled usage
One access point for all approved models, with permissions, budgets and clear rules instead of shadow AI.
Demonstrably compliant
Logs, policy and risk assessment document responsible use.
Our services in detail
- Privacy filter: automatic detection and pseudonymisation of names, addresses, IBANs, customer and contract data
- Central AI gateway with model approval, permissions per department and cost control
- Protection against prompt injection and data leakage for agents and knowledge assistants
- Logging and analysis of AI usage for audit and data protection
- AI policy, risk assessment and documentation under the EU AI Act
How it works
- 1Inventory
Which AI services are used today, with what data and by whom?
- 2Define rules
Together we define which data may go where and which models are approved.
- 3Technical implementation
The AI gateway, privacy filter and sign-in via your user management are set up.
- 4Monitoring
Ongoing analysis, adjustment of the filters and regular review.
Frequently asked questions
What exactly does a privacy filter do?
It detects personal data, or data defined as confidential, in texts and documents, such as names, addresses, account numbers or customer numbers, and replaces it with placeholders before the request goes to an external model. The original values are restored in the response.
Is it not enough to ban ChatGPT in the company?
In our experience, employees then switch to private accounts, which increases the risk. A secure, approved access point with safeguards is more effective than a ban.
What is prompt injection?
An attacker hides instructions in an email, web page or document that are meant to trick an AI into disclosing data or performing the wrong actions. We limit permissions, check inputs and monitor agents to reduce this risk.
What does the EU AI Act require of us?
Since February 2025, companies have had to ensure that employees who use AI systems have sufficient AI literacy. Depending on the use, transparency and documentation obligations may apply. We help with classification, policy and documentation.
Related services
More services in the area of AI & Automation that complement AI Security & Privacy Filter.
AI Consulting & Strategy
From the first use case to the roadmap: we show you where AI measurably saves time in your business, and what you can safely skip.
Learn more →
Local AI & Private AI
Powerful language models on your own hardware, on site or in our data centre in Germany. Your data never leaves your control.
Learn more →
RAG & AI Knowledge Assistants
An AI assistant that knows your manuals, contracts and tickets, with a source for every answer and respect for access rights.
Learn more →Let us talk about AI Security & Privacy Filter.
Call us or write to us. We will get back to you promptly and discuss what you really need.
Or call us directly: +49 821 66 0 99 0