AI security: use artificial intelligence safely
Azure AI Content Safety
GDPR-compliant AI use
AI security in detail
Content filtering
Harmful or inappropriate content is detected and automatically blocked before it can cause damage—including phishing attempts, malware links, and content that violates internal policies or legal requirements.
Data protection for AI tools
Specifically, we secure local AI systems for social institutions that work with highly sensitive personal data—with multi-level security concepts audited at defined intervals by independent experts to ensure no data leaves the organization.
Access control
Only authorized employees get access to sensitive AI applications and functions—with clearly defined roles and permissions aligned with your existing user and access management.
Monitoring & reporting
Use of AI services is logged and regularly evaluated—for full transparency on which data was processed when and by whom, including alerts for suspicious behavior.
Secure, local AI for a highly sensitive area
For a social services provider working, among other things, in child and youth protection and in psychological counseling for vulnerable people, a domain-specific AI was to help with drafting texts. The original approach by an external consultant envisioned a cloud-based AI combined with a user-based billing portal—meaning data leakage to third countries could not be reliably ruled out, despite planned anonymization.
Given the highly sensitive data in this area, that was too risky for us. Instead, we fully isolated the solution behind a Sophos XGS firewall and deployed a local AI model on a dedicated server with a dedicated GPU. Since then, it has been running reliably in production, with no possibility of data leaking externally.
Make company knowledge usable for AI—under control
A secure Retrieval-Augmented Generation (RAG) system does not adopt information unfiltered. It separates selection, review, and use: per request, the AI receives only approved, necessary excerpts and provides traceable references to its sources.
Select
Only technically relevant and explicitly approved sources enter the process.
Quarantine
New content remains separate until origin, purpose, and protection requirements are clarified.
Clean up
Personal, confidential, and unnecessary information is removed or minimized.
Review
A human checks content, rights, timeliness, and technical suitability.
Approve
Only confirmed knowledge is versioned and published for search.
Test
Quality, source traceability, knowledge gaps, and unwanted answers are continuously reviewed.
Sources, rights, and responsibility remain visible
- Verifiable answers: Source, version, and location remain traceable.
- No source, no binding statement: Knowledge gaps and contradictions are clearly flagged.
- Rights per request: Directory and communication data is included only in the permitted context.
- Human approval: AI suggestions are only added back to the knowledge base after review.
- Data sovereignty: Local or clearly scoped hybrid operation prevents uncontrolled leakage.
AI use that takes security and privacy seriously
Fully managed IT infrastructure—proactive, predictable, and secure.
IT infrastructure
All IT areas are interconnected and optimally aligned—for a stable, secure, and scalable infrastructure.
Microsoft 365
Microsoft 365 setup and management—secure, GDPR-compliant, and protected against cyberattacks.
Cloud: private, hybrid, or public
Scalable cloud infrastructures for private, hybrid, and public cloud—flexible, secure, and sovereign.
Use AI securely and in line with GDPR
Contact us
Gabler Systemtechnik GmbH
Scheidegger Strasse 8
81476 Munich
Get started with us!
May our employees use ChatGPT and similar tools?
Grundsätzlich ja, aber mit klaren Regeln: Welche Daten dürfen eingegeben werden, welche Tools sind freigegeben und wie wird die Nutzung protokolliert. Wir helfen Ihnen, diese Regeln festzulegen und technisch durchzusetzen.
What is Azure AI Content Safety?
Ein Dienst, der automatisch schädliche, beleidigende oder unangemessene Inhalte in Texten und Bildern erkennt und filtert – nützlich etwa für Webportale mit nutzergenerierten Inhalten oder KI-gestützte Chat-Anwendungen.
Is the use of AI tools compatible with the GDPR?
Das hängt stark davon ab, welches Tool wie eingesetzt wird und welche Daten verarbeitet werden. Wir prüfen Ihre konkreten Anwendungsfälle und richten Zugriffskontrollen sowie Datenflüsse so ein, dass der Einsatz GDPR-compliant bleibt.