Trust & Compliance
How we build AI responsibly, prepare for the EU AI Act, protect your data, keep systems secure, handle intellectual property and support you after delivery.
Last updated: 4 November 2026
01
Responsible AI policy
Every AI system we build follows these principles, from scoping to maintenance.
- Human oversight
- Our systems support decisions; people make them. Operators can review, correct or reject every output. We do not implement autonomous closed-loop control unless the client explicitly decides so after a separate risk assessment.
- Fit for purpose and validated
- Models are evaluated on representative data held out from training, with error analysis by class, camera, shift or operating condition. We report limits as clearly as results.
- Transparency and explainability
- Users see scores, alternatives and the factors behind a prediction. Each delivered model comes with documentation of its purpose, data, evaluation and known limits.
- Fairness
- Where a system affects people, we check data and results for bias. We do not build systems for practices prohibited by the EU AI Act, such as social scoring or manipulative techniques.
- Data minimisation and privacy
- We use only the data a task needs and anonymise or pseudonymise personal data whenever possible.
- Monitoring
- After deployment we monitor performance and retrain when products, data or conditions change.
- Accountability
- Every project has a named person at CIVINTEL responsible for these commitments.
02
EU AI Act readiness
We help clients understand where their AI systems stand under Regulation (EU) 2024/1689.
- Risk classification at scoping
- Each system is classified against the AI Act categories (prohibited, high-risk, transparency obligations, minimal risk) before development starts, and the result is recorded in the proposal.
- Typical industrial use
- Quality inspection, energy optimisation and formulation support usually fall outside the high-risk categories unless the system acts as a safety component of a product or machine. We check this case by case.
- If a system is high-risk
- We tell you early and plan the required risk management, data governance, technical documentation, logging, human oversight and accuracy testing.
- Documentation you receive
- Intended purpose, data sources, training and evaluation methods, performance, limits and instructions for use, whatever the risk class.
- AI literacy
- We train the people who will use or supervise the systems, in line with Article 4 of the AI Act.
03
Client data & confidentiality
Your data and documents remain yours.
- Confidentiality first
- We sign a non-disclosure agreement before receiving any data or documents.
- Data processing agreement
- When we process personal data on your behalf, we sign a data processing agreement under Article 28 GDPR. A template is available on request.
- Where data lives
- On your premises or in EU-hosted environments. We do not transfer client data outside the EU/EEA.
- Purpose limitation
- Your data is used only for your project. We never use it to train models for other clients.
- Restricted access
- Only the named CIVINTEL team members working on your project have access.
- End of project
- Data is returned or deleted, with written confirmation, unless we agree otherwise in writing.
- Subcontractors
- Only with your prior written approval and under the same obligations.
04
Information security
Proportionate security controls for a small engineering company that handles industrial data.
- Encryption
- Encrypted storage on company devices and encrypted transfer (TLS) for all data exchanges.
- Accounts and access
- Individual company accounts with multi-factor authentication, and least-privilege access to client systems and data.
- Sensitive environments
- Fully on-premise or offline deployment is possible when data must not leave the plant.
- Secure development
- Version control, code review, dependency updates and separate test and production environments.
- Incident handling
- If an incident affects your data, we notify you without undue delay and in any case within 72 hours of becoming aware of it, then investigate, contain and report on remediation.
05
Intellectual property & licensing
Clear ownership from the start, written into every contract.
- Your project deliverables are yours
- Models trained on your data, code written specifically for you, reports, drawings and results belong to you upon full payment.
- Our background tools
- CIVINTEL keeps its pre-existing tools, libraries, methods and know-how, and grants you a perpetual, non-exclusive licence to use them as part of your deliverables.
- Your data stays yours
- We acquire no rights over your data.
- Open source
- Open-source components are listed in the documentation with their licences.
- References
- We mention a client or project publicly only with the client’s permission, and without confidential details.
- Engineering studies
- Studies and drawings are protected by copyright under Greek law; the client may use them for the project they were prepared for.
06
Support & service levels
How to reach us and how quickly we respond.
- Channels
- Email, phone and video call, Monday to Friday, 09:00–17:00 (Greek time).
- Support contracts
- Service levels, response times and, for systems we host, availability targets are written into each support contract.
| Request | First response |
|---|---|
| General enquiries and new projects | Within 2 working days |
| Clients with a support contract | Within 1 working day |
Questions about these commitments, or need a signed NDA or data processing agreement? Contact · kliakos@civintelgr.com
Have a question your data or your building could answer?
A first conversation is free and without obligation. Tell us what you want to achieve and we will tell you honestly whether and how we can help.