AI Policy

This AI Policy explains how the AI assistant in intakto (“Hermes”, referred to as “she”) works, what she can and cannot do, and how we keep you in control. It forms part of the Terms of Service and should be read with the Privacy Policy.

1. What Hermes is

Hermes is software built on artificial intelligence and large language models. She uses your persistent memory and your instructions to generate responses and to carry out tasks you approve. To do this, the Service sends relevant content to an AI model provider for processing; see the Privacy Policy and the Sub-processor List.

2. The limits of AI output

2.1. AI output is generated by a probabilistic model. It may be incorrect, incomplete, out of date, biased, or fabricated (“hallucinations”), even when it sounds confident.

2.2. Hermes is not professional advice. She does not provide legal, medical, financial, tax, accounting, psychological, or other regulated advice and does not replace a qualified professional.

2.3. You must review and verify outputs before relying on them. You are responsible for decisions you make using outputs.

3. Human-in-the-loop and approval

3.1. We design the Service so that you stay the decision-maker. Meaningful actions — for example sending a message or making a change in a connected service — are designed to be presented for your review and approval rather than taken silently.

3.2. You choose what to instruct and what to approve. Where you grant standing instructions or routines, you remain responsible for the actions carried out under them, and you can change or revoke them.

3.3. We keep an action/audit log so that what was done on your behalf is traceable.

4. Bounds of agentic actions

4.1. Hermes acts only within the access you grant and the connected services you authorise. She does not act outside that scope.

4.2. You are responsible for the consequences of actions you authorise, including in connected services. See the Terms of Service (Section 7) and the Acceptable Use Policy.

5. Your data and model training

5.1. We do not use the content of your memory or your conversations to train AI models, and we contract with our model provider so that your content is not used to train their models.

5.2. If any future feature would rely on such use, we would ask for your separate, opt-in consent first.

6. Automated decision-making

We do not use the Service to make solely automated decisions that produce legal or similarly significant effects on you. Where you use Hermes to support decisions about other people, you must keep appropriate human oversight and a lawful basis.

7. Reporting problems

If an output or action causes a problem, or you see misuse, tell us at hello@intakto.eu so we can help and improve the Service.

8. Contact

HumanOps B.V., Amsterdam, 1091EE, Netherlands. Email: hello@intakto.eu. KvK number: [KvK number - to be added]. VAT/BTW number: [BTW number - to be added].

This document is provided for information and is not legal advice.