Security & responsible AI
Defensive engineering with explicit scope, human responsibility and verifiable outcomes.
What we do
Our core work is software development and authorised penetration testing of applications, APIs and infrastructure. We also work with Wazuh deployment at scale, detection engineering, hardening and vulnerability remediation. Our engineering practice spans Kubernetes, Linux, virtualisation and CI/CD.
Authorisation comes first
Security assessment work requires a defined scope and permission from the system owner. Targets, permitted actions and operating constraints are agreed before work starts. Controlled test environments are used where appropriate to validate findings without unnecessarily affecting production.
How we approach AI-assisted work
We have used AI in our engineering work for years. In security workflows, AI can support analysis of redacted alerts, detection-rule development, code review and remediation proposals. Findings need supporting evidence; significant changes require human judgement and validation.
Practical AI in defensive engineering
Useful applications include reviewing maintained code, improving detection rules with redacted telemetry, analysing infrastructure configuration and testing proposed fixes. We keep the task bounded, check the evidence and validate consequential changes before they reach production.
AI supports the engineer’s work. It does not replace authorisation, sound judgement or responsibility for the outcome.
Data & operational controls
- Limit inputs to information necessary for the task; remove secrets and identifying details where possible.
- Use scoped access and separate approved workflows from public-facing traffic.
- Review findings before acting and validate changes with appropriate tests.
- Respect provider retention terms and customer data obligations; do not assume zero data retention.
Report a concern
For a security concern affecting this website or infrastructure we operate, contact info@uhlhosting.ch with a brief description and a safe way to reach you. Please do not include passwords, personal data or exploit payloads in the initial message. We can agree an appropriate channel for details.