AI supplementary terms

Revision: 25 September 2026. Contractual terms apply only when incorporated into the relevant agreement; this revision does not retrospectively change existing contracts.

1. What these terms cover

These terms supplement the service agreement only for the AI features enabled in the customer’s order. The product, selected model, tool connections and deployment determine what data is processed and which actions are possible. A research result or demonstration does not warrant performance on every task.

2. Inputs, outputs and permitted use

You retain rights in your inputs. As between you and Mindtro, you may use outputs for the agreed purposes; Mindtro assigns any rights it holds in those outputs to the extent legally possible. This does not guarantee that output qualifies for copyright, is unique, is accurate or is free of third-party rights. You must have the right to submit the inputs and assess output before consequential use.

3. Customer content and training

By default, Mindtro does not use customer inputs or outputs to train or fine-tune general models. Any optional contribution requires a separate, specific agreement describing the data, purpose, safeguards and withdrawal or termination process; accepting general service terms is not that agreement. Service diagnostics must be limited to what is necessary and governed by the applicable processing terms.

Before enabling a model provider, its processing, retention and training terms must be checked for the chosen account and service. We do not promise a provider’s zero retention or no-training setting merely from its brand name. Customer-managed models or API credentials remain subject to the agreed integration boundaries.

4. External tools and human decisions

A customer must authorise connected systems, tool permissions and any external side effects. Evaluate the required level of human review for financial, legal, employment, safety or other consequential decisions. An AI recommendation alone is not approval to place an order, change a production system or grant access. Configure and verify the applicable permission and approval controls before use.

Security testing must be limited to authorised targets. Findings indicate the tested scope and do not certify the complete absence of vulnerabilities. Customers must not use AI features unlawfully or circumvent restrictions and safety controls.

5. Providers, data protection and changes

The selected model and tool providers are identified in the order or service-specific schedule before covered data is sent. Only terms actually relevant to an enabled provider are incorporated and made available. A list of potential integrations is not a list of active subprocessors. Personal-data processing follows the DPA, provider schedule and privacy notice.

Material changes to a model, provider or processing purpose must follow the agreement and any applicable subprocessor notice or consent process. No model change silently enlarges training permissions.

6. Quality, security and liability

Outputs may contain errors, omissions or bias. Validate fitness for the intended workflow and retain human responsibility where required. This explanation does not remove agreed functionality, statutory conformity rights or remedies. The service terms govern liability, including their mandatory-rights exceptions. Certification information is limited to the entity and scope stated in the security policy.

Questions, notices and requests: legal@mindtro.com. General enquiries: info@mindtro.com, +49 30 166381616.