AI Principles
Version 1.0 as of 08 November 2026
We use AI-assisted tools to work more efficiently. The following principles apply to every client project:
1. People stay accountable
AI supports our work, but it never makes the decisions. Every deliverable is reviewed by a qualified expert, checked against primary sources and approved before release.
2. No real client data goes into AI
We never share real client data with AI tools – no names, addresses or contact details, no company, product or trade names, no site data and no other identifying information.
3. Consistent anonymisation
Client-specific content is always anonymised or pseudonymised before any AI tool is used. Indirect identifiers such as unusual formulations, specific processes, regions or dates are also abstracted, so that the client cannot reasonably be identified. Mapping back to the real data happens only on our side, outside any AI tool.
4. Memory features under control
Memory and history features of AI tools are controlled or switched off for client projects. As no real client data ever enters AI tools, it cannot be used to train AI models.
5. Our clients decide
Clients can exclude the use of AI for their project in full or in part, rule out specific tools, request that AI-assisted content be labelled, or require that AI tools are used only in a mode without history or memory.
6. AI literacy
In line with Article 4 of Regulation (EU) 2024/1689, everyone who works with AI on our behalf is trained in how it works, its limitations, data protection and anonymisation.
The full AI policy forms part of our proposals. If you have any questions, please send your request via the SafetyCampus ticket system or contact us at
