AI Usage Policy
Artificial intelligence is a real part of our toolkit — it supports our research, content preparation and work with code. This page describes openly how we use it: which tools we work with, how we protect client data, and who is accountable for the results.
1. Tools — chosen deliberately
We use professional AI tools — in particular Claude models by Anthropic, including Claude Code for programming work — exclusively in configurations where the data we enter is not used to train the models (data sharing deliberately disabled). We do not use free tools that learn from user input when working with client materials. We do not circumvent the safeguards of tools or accounts — if a tool protects something for security reasons, we respect that.
2. Client data — a hard line
The following never reach AI tools:
- passwords, API keys and access credentials (SSH, FTP, admin panels),
- personal data of our clients' customers,
- confidential documents, unpublished contracts and financial data.
We minimise and anonymise working materials — we pass on only what is necessary for the task at hand. When in doubt whether something can be shared — we don't share it. Convenience never wins over the security of client data.
3. A human decides and is accountable
AI speeds up the work and suggests solutions; the choice, the review and the responsibility stay with us. No text, code or recommendation reaches a client without human review. Changes to production environments and operations on client data are performed only after human approval.
4. No fabricated data
Client materials and our own publications contain no invented figures, reviews or promises. Facts, metrics and prices are only ever stated when confirmed — and where data is missing, we say so plainly.
5. Honesty with clients — including under NDAs
We attribute AI co-authorship where appropriate (for example in code change history). Clients who sign non-disclosure agreements (NDAs) with us are informed before the engagement starts that the work may be supported by AI tools and that working project materials may be processed by the tool provider's infrastructure — in our configuration, without the data being used to train models. The scope of permitted processing should be clear to both parties before any agreement is signed.
6. Compliance and updates
We keep this policy consistent with applicable regulations (GDPR, the EU AI Act) and update it whenever our tools or the law change. Editorial and substantive responsibility for everything we publish and deliver remains with people.
For a practical guide on setting up an AI policy in your own company, see: Company AI policy — where to start. Questions about our policy: info@invisio.digital.
Last updated: 12 August 2026.