AI Usage Policy

1. Why this page exists

The EU AI Act (Article 50) requires businesses to be transparent about the AI systems they use. I run SevinHub from Belgium, so this applies to me directly, and I would rather over-share than hide it in fine print. This page is my policy, my inventory summary, and my commitment in one place.

2. Where AI is used on this site and in my products

In EU AI Act terms, SevinHub acts as a Deployer of third-party General Purpose AI (GPAI) models. I integrate models from providers such as OpenAI and Anthropic through their APIs; I do not train or provide my own foundation models. My obligations here are the transparency and oversight duties that apply to a deployer.

• Blog content: some articles are drafted with AI assistance. Every AI-assisted post carries a visible "AI Assisted (human reviewed)" label. I review and edit everything before it is published.
• Images: some illustrations and blog images are AI-generated. I am rolling out machine-readable AI markers (IPTC digital source type metadata) for these images.
• Customer support: replies are written or reviewed by me personally. I sometimes use AI tools to draft responses faster, but a human reads and approves every reply before it reaches you.
• Products: some of my products contain AI features (for example the DoAnything app builder and the ActHub AI drafter). Those features are clearly presented as AI at the point of use, and they only run when you actively use them, with your own or my disclosed configuration.
• Development: I use AI coding assistants internally to build and maintain my software. Everything shipped is reviewed by me.

Bring Your Own Key (BYOK). Where a product lets you use your own API keys (for example the ActHub AI drafter with your own OpenAI or Anthropic key), your prompts and their responses are processed under your direct agreement with that provider. In that mode I do not log, store, or reuse those prompts on my central infrastructure; the data-processing terms are governed by your contract with the model provider, not by me.

3. What I do NOT do

• No emotion recognition, no biometric categorisation, no social scoring, no deepfakes.
• No AI decisions about people (no automated screening, scoring, or profiling of customers).
• No selling or sharing of your data with AI vendors beyond what is needed to run the feature you are using.

4. Human oversight

Every AI output that reaches customers (content, support replies, product changes) passes through human review. I am a registered self-employed professional and a security practitioner; I treat AI as a tool under my responsibility, not as a decision maker.

5. If something looks wrong

If you believe AI-generated content on this site is inaccurate, unlabelled, or harmful, email contact@sevinhub.com. I aim to respond within 72 hours and will correct or remove confirmed issues. You can always reach a human: it is just me and my team, and we answer personally.

This 72-hour target is my informal internal mechanism for putting things right quickly. It does not replace your legal rights: if you believe there is a systematic compliance failure, you retain the right to lodge a formal complaint with the competent Belgian AI supervisory authority (BIAM, the Belgian Institute for AI / market-surveillance authority designated under the EU AI Act).

6. Review rhythm

I review this policy and my AI inventory quarterly, and whenever I add a new AI system. The full machine-generated inventory and transparency report is maintained with my own compliance product, ActHub, and the public version is available on my AI transparency page.