AI Transparency Notice (EU AI Act)

Tamanor includes an AI Risk Engine that analyses publicly available content (comments, reviews, mentions) retrieved via official platform APIs and assigns a risk level, one or more risk categories, and a sentiment, together with a confidence score and a short rationale. Its purpose is to help our Customers prioritise and triage moderation — not to replace human judgement.

Tamanor's AI Risk Engine is a content-triage / classification tool. It is not a prohibited practice under Article 5 and is not, in our assessment, a high-risk system under Annex III of the AI Act (it does not, for example, determine access to essential services, employment, credit, or law-enforcement outcomes). We treat it under the AI Act's transparency regime and apply human-oversight safeguards voluntarily as good practice. Should the classification change with product evolution, we will reassess and update this Notice.

Consistent with Article 50, the Service makes clear within the interface which outputs are AI-generated or AI-assisted (e.g. risk labels, sentiment, suggested priority, and any suggested reply text are marked as machine-produced). Users always see that a classification came from the AI Risk Engine, including the engine version and confidence, so they are never left to assume a human made the assessment.

The system is designed for human-in-the-loop operation:

• Sensitive categories (e.g. legal threats, self-harm) and high/critical severity items are always routed to a person for review and approval; rules cannot override this.
• Automated actions (such as hiding a comment) are permitted only when confidence is above a high threshold, the platform's official API supports the action, and the category is not sensitive — otherwise the item waits for human approval.
• Every automated and human action is recorded in an immutable audit log (who/what/when/why, engine version, confidence).
• The design is fail-safe: on any uncertainty, the system does nothing and surfaces the item for human review rather than acting.

We therefore do not use AI to make solely automated decisions producing legal or similarly significant effects on individuals within the meaning of Article 22 GDPR.

The AI Risk Engine processes the content and limited author metadata already lawfully retrieved on the Customer's behalf via official APIs. It does not scrape data, does not use platform passwords, and (where an external AI provider is enabled by the Customer) processes data under a data-processing contract and the safeguards in our DPA and Subprocessor List. Where no external provider is configured, classification runs on our own rules/engine.

Automated classification can be wrong: it may mislabel tone, miss context, sarcasm, dialect or minority-language nuance, or reflect the limitations of its training. Outputs are decision-support, not verdicts. Customers remain responsible for their moderation decisions and should apply human judgement, especially for edge cases and protected expression. Tamanor does not automatically treat a low rating or criticism as risky.

You may ask for information about the logic involved and request human review of any AI-assisted outcome that affects you. For content processed on a Customer's behalf, direct such requests to that Customer (the controller); we will assist them. For our own processing, contact info@tamanor.com. See also the Privacy Policy and Data Subject Rights Policy.

We monitor developments under the AI Act (including guidance, harmonised standards and the Code of Practice) and will update our practices and this Notice accordingly, including any machine-readable marking obligations for AI-generated content as technical standards mature.