Acceptable Use Policy (AI)

This is an English translation provided for convenience. The legally binding version is the Czech original; in case of any discrepancy, the Czech version prevails.

Acceptable Use Policy (AUP)

Publisher: Digital D&A s.r.o. | Company ID: 23691271

Seat: Školská 660/3, Nové Město, 110 00 Prague

Contact: support@awentail.com | awentail.com

Applies to: all customers and users of the Awentail platform

Effective from: 14 March 2026

1. Purpose and scope

This Acceptable Use Policy (the “AUP”) sets out the rules for the responsible, lawful and ethical use of the Awentail platform and its AI assistants. The AUP supplements the Terms of Service and the Privacy Policy and, in case of conflict, prevails over them in areas concerning the use of AI.

The AUP applies to:

• Awentail customers (businesses and individuals registered on the platform) — when configuring, training and deploying AI assistants;
• visitors of customers’ websites — when interacting with AI assistants deployed by Awentail customers;
• customers’ employees and collaborators — when managing and monitoring AI assistants.

The aim of the AUP is to ensure that Awentail AI assistants deliver value without causing harm to users, third parties or society as a whole.

2. Acceptable use of the AI assistant

The Awentail platform is designed for legitimate business purposes. Acceptable and supported uses include:

✓ The platform is fully optimised for the above purposes, and the customer is responsible for the correct configuration of the AI assistant in accordance with this AUP.

3. Prohibited use

⛔ The following categories of use are strictly prohibited. A breach may lead to immediate cancellation of the account without compensation and to the customer’s legal liability.

3.1 Practices prohibited by the EU AI Act (Art. 5, effective from 2 February 2025)

In accordance with Regulation (EU) 2024/1689 (EU AI Act), it is prohibited to use the Awentail AI assistant for:

• Subliminal manipulation — techniques that operate beyond a user’s conscious perception and influence their behaviour in a way that may cause them harm.
• Exploitation of vulnerabilities — targeted exploitation of the vulnerabilities of specific groups (children, the elderly, persons with mental disabilities, economically vulnerable persons) in a way that may cause them harm.
• Social scoring — evaluating or classifying natural persons based on their social behaviour, leading to their unjustified detriment or discrimination.
• Biometric identification — real-time identification of natural persons in public spaces without a legal basis.
• Emotion recognition — recognising emotions in a workplace or educational setting without a legal basis.
• Predictive policing — predicting the risk of committing a crime based on the profiling of a natural person without objective grounds.

3.2 Deception and manipulation

• Passing the AI assistant off as a real natural person in a way that actively deceives users.
• Spreading deliberately false, misleading or fraudulent information.
• Creating phishing scenarios or impersonating trusted institutions (banks, authorities, healthcare facilities) for the purpose of fraud.
• Psychological manipulation of users to force decisions they would not otherwise make.

3.3 Illegal content and activities

• Distributing content that incites hatred, discrimination or violence based on race, gender, religion, nationality, sexual orientation or other protected characteristics.
• Generating or distributing child sexual abuse material (CSAM) or content that sexualises minors.
• Facilitating illegal activities, including fraud, money laundering, human trafficking or drug trafficking.
• Infringing intellectual property rights — systematically reproducing protected copyrighted works without authorisation.
• Collecting or processing personal data without a legal basis or in breach of GDPR.

3.4 Misuse of technology and infrastructure

• Attempting to breach, overload or circumvent the platform’s security measures.
• Reverse engineering, decompiling or extracting AI models, prompts or the internal logic of the Platform.
• Automated mass querying (scraping) for the purpose of extracting data or testing system limits.
• Benchmark testing or analysing the Platform for the purpose of developing a competing product.
• Sharing access credentials or reselling access to the Platform to third parties.
• Using the Platform in a way that may jeopardise the stability, performance or availability of the service for other customers — including excessive load on the infrastructure or circumventing rate limits.

3.5 Critical sectors without appropriate safeguards

Without ensuring appropriate human oversight and without express consultation with Awentail, the AI assistant must not be the primary tool for:

• Medical diagnosis or treatment recommendations — the AI assistant may provide general information, not medical advice.
• Automated legal decision-making with legally binding effects — the AI may explain concepts, not provide legal advice.
• Financial advice or investment recommendations — the AI may inform, not recommend specific financial products.
• Assessing creditworthiness or making insurance decisions without human oversight.
• Recruitment — automated evaluation or rejection of candidates without human review (high-risk AI under Annex III of the AI Act).

4. Customer obligations when deploying an AI assistant

4.1 Transparency towards users (Art. 50 of the EU AI Act)

The customer is obliged to ensure that visitors of their website are clearly informed that they are communicating with an AI system and not a human person. This information must be provided no later than at the start of the first interaction.

⚠ Hiding the AI nature of the assistant or actively passing it off as a human is prohibited and constitutes a breach of Art. 50 of the EU AI Act and of this AUP. The customer bears sole responsibility for meeting this obligation.

4.2 Knowledge Base configuration and content

• The customer is responsible for the accuracy, timeliness and lawfulness of all content uploaded to the Knowledge Base.
• The customer must not upload third parties’ personal data without their consent or another legal basis.
• The customer must not upload special categories of personal data (health, biometric, genetic, etc.) without prior consultation with Awentail.
• The customer regularly updates the Knowledge Base so that the AI assistant provides accurate and up-to-date information.

4.3 Human oversight

The customer ensures appropriate and ongoing human oversight of the AI assistant’s operation:

• Regularly monitors conversations, outputs and captured contacts.
• Establishes an internal escalation procedure — when and how a human operator should join or take over a conversation.
• Maintains a functioning mechanism to immediately and temporarily disable or restrict the AI assistant in the event of anomalous behaviour.
• Ensures that no automated decision with legal or significant effects is executed without human review.

4.4 Staff AI literacy

The customer ensures that employees working with the Platform or with the AI assistant’s outputs complete appropriate basic training covering:

• An understanding of how the AI assistant works and the limits of generative AI (hallucinations, bias, sensitivity to input data).
• Knowledge of the obligations under the EU AI Act and GDPR in the context of deploying an AI assistant.
• Procedures for identifying and reporting incidents or inappropriate AI behaviour.

5. Rules for working with the AI assistant’s outputs

5.1 Limits on the reliability of outputs

The AI assistant generates responses based on probabilistic models. The AI assistant’s outputs:

• may contain inaccuracies, outdated information or hallucinations (convincing-sounding but incorrect statements);
• do not constitute legal, financial, medical or other professional advice;
• must not be presented as guaranteed correct or as a substitute for professional human judgement.
• are provided without any warranty of accuracy or suitability for a particular purpose — Awentail provides no warranty, express or implied, regarding outputs generated by AI models.

ℹ The customer is obliged to reflect this in the AI assistant’s configuration — for example, through a system prompt that instructs the AI assistant to recommend, in relevant cases, that the user contact an expert or a human company representative.

5.2 Responsibility for outputs towards users

The customer, as the website operator, is responsible for how the AI assistant’s outputs are presented to its customers. The customer must not:

• present the AI assistant’s outputs as its own expert opinion or as warranties;
• rely on AI outputs in decisions that could cause serious harm (health, financial, legal) without those decisions passing through human review;
• distribute AI outputs about third parties without their consent, in particular where personal or sensitive information is concerned.

6. Classification of deployment scenarios by level of risk

Below is an indicative overview of typical Awentail AI assistant deployment scenarios and their classification from the perspective of the EU AI Act and this AUP:

This classification is indicative and does not cover all possible scenarios. The customer is obliged to assess the risks of their specific deployment independently. In case of doubt, contact support@awentail.com.

7. Protection of vulnerable groups

7.1 Children and minors

If the customer operates a website that may be visited by minors, they are obliged to take appropriate measures to protect them:

• Not to configure the AI assistant in a way that could put minors at risk or expose them to inappropriate content.
• Not to obtain personal data from minors without the verifiable consent of parents or legal guardians.
• Not to use persuasive or manipulative techniques towards minors.

7.2 Persons in a vulnerable situation

The AI assistant must not be configured to deliberately exploit the vulnerability of persons in a difficult life situation (financial distress, health crisis, mental instability) to achieve the customer’s business goals.

The customer is obliged to set up an escalation mechanism for cases where the AI assistant detects or receives a message suggesting that the user needs immediate help (a crisis situation, danger to health).

8. Reporting incidents and anomalies

The customer is obliged to inform Awentail without delay at support@awentail.com if:

• The AI assistant starts generating content that breaches this AUP (hateful content, deceptive responses, prohibited practices).
• The customer finds that the AI assistant has been abused by a third party for unauthorised or harmful purposes.
• A security incident or data breach occurs in connection with the operation of the AI assistant.
• The customer receives a complaint from a user suggesting that the AI assistant caused harm.

After receiving a report, Awentail investigates and informs the customer of the measures taken. In the meantime, the customer temporarily restricts or deactivates the AI assistant where technically possible and where the situation requires it.

9. Consequences of breaching the AUP

The customer bears sole responsibility for damage caused by a breach of this AUP — towards Awentail, third parties and regulatory authorities. Awentail is not liable for a customer’s conduct in breach of this AUP.

10. Awentail’s commitments

For its part, Awentail undertakes to operate the AI platform responsibly:

• Technically block configurations that would enable practices prohibited under Art. 5 of the EU AI Act.
• Monitor use of the Platform, both automatically and manually, in order to detect breaches of this AUP, protect the integrity of the infrastructure and ensure security for all customers. The results of monitoring may serve as a basis for measures under Art. 9.
• Update this AUP in line with developments in the EU AI Act and EU AI Office guidance.
• Provide customers with documentation, recommended practices and system-prompt templates for responsible deployment.
• Communicate transparently about the third-party AI models used and any of their limitations.
• Not use customers’ data to train shared AI models without express consent.
• Transparently disclose that the Platform uses third-party AI models. Awentail is not liable for the content generated by these models or for their technical limitations, errors or availability.

10a. Exemptions from this Policy

Awentail may, in individual cases, grant a written exemption from a specific restriction of this AUP, provided the proposed use complies with applicable law, does not harm third parties, and the customer demonstrates sufficient safeguards for responsible deployment. Please send exemption requests to support@awentail.com.

11. Updates to this Policy

Awentail is entitled to update this AUP in line with developments in legislation (in particular the EU AI Act and GDPR), technology and business needs. The customer will be informed of material changes by e-mail at least 14 days in advance.

The current version of the AUP is always available at awentail.com/acceptable-use-policy.

By continuing to use the Platform after the updated AUP takes effect, the customer expresses agreement with the new wording.

12. Contact

Please send questions, incident reports or exemption requests regarding this Policy to:

support@awentail.com | Digital D&A s.r.o., Školská 660/3, 110 00 Prague | awentail.com

Digital D&A s.r.o. | Company ID: 23691271 | AI Use Policy v1.0 | Effective from 14 March 2026