Responsible AI.
Our commitments.

01Classification under the AI Act

The MyKonci Service uses generative artificial intelligence models to produce content (conversational replies) and analysis systems to produce scores and recommendations (Cortex).

Within the meaning of Regulation (EU) 2024/1689 (the "AI Act"), the Service is classified as follows:

• Limited-risk AI system (Title IV of the AI Act), chatbot and voice agent: transparency obligation (inform the user they are interacting with an AI), no other substantial obligations.
• Limited-risk AI system, Cortex: analysis and recommendations without automated decisions producing legal or significant effects within the meaning of article 22 of the GDPR.

MyKonci does not deploy high-risk or unacceptable-risk AI systems within the meaning of the AI Act.

02Transparency towards users

In line with article 50 of Regulation (EU) 2024/1689 (AI Act), guests interacting with the Service must be informed that they are interacting with an AI system. This principle is implemented as follows:

• Information by the customer, as controller: it must inform its guests in line with articles 13 and 14 of the GDPR and tell them they are interacting with an AI system, for example in its rental listings, welcome messages or arrival instructions.
• Voice identification: the voice agent explicitly introduces itself as an automated assistant when answering calls. Audio recordings generated are identifiable as outputs of an AI system.
• Written identification: the chatbot's first message can include a notice of its automated nature, configurable by the Customer.

Failure by the Customer to comply with this information obligation is a breach of this charter: like any breach (article 05), it may lead to suspension of the Service.

03Human oversight and decisions

MyKonci applies the principle of human oversight:

• No decision producing legal or significant effects within the meaning of article 22 of the GDPR is taken in a fully automated way regarding guests.
• Cortex produces recommendations. During the observation phase, it sends no message, neither to guests nor to the customer's team. It can only send a message itself if the owner of the customer's concierge business has enabled the "Let Cortex act" option.
• Chatbot and voice-agent replies rely solely on the information provided by the customer; in case of missing or ambiguous information, the AI escalates to a human rather than making things up.
• The customer can schedule the automatic sending of certain messages; the conversational assistant then replies to guests on its own, without prior human validation of each message, unless review mode is enabled (in which every reply awaits an operator's validation before sending). None of these automations amount to a decision within the meaning of article 22 of the GDPR. A customer who enables an automation feature becomes the controller within the meaning of the GDPR and must provide an opt-out mechanism for guests.

04Data quality and bias

MyKonci implements measures to limit the biases and errors inherent to AI:

• Strict prompt framing (instruction system) limiting the AI's scope of action.
• Escalation mechanism to a human in case of doubt or missing information.
• Strict isolation between customers: no customer's data influences another's analyses.
• Periodic quality audits of replies and recommendations.
• Customers can report an error; corrections are integrated into the service improvement process.

05Prohibited uses (Acceptable Use Policy)

The Customer undertakes not to use the Service to:

• Produce or distribute illegal, defamatory, hateful, discriminatory content or content contrary to public order.
• Manipulate guest behaviour in a deceptive, hidden way or in a manner harmful to their free consent.
• Rate or categorise people on the basis of protected characteristics (origin, religion, sexual orientation, political opinions, health, etc.) within the meaning of article 5 of the AI Act.
• Set up a general social scoring of guests.
• Extract facial images or biometric data for identification purposes.
• Infer people's emotional state in the workplace or educational settings within the meaning of article 5.1.f of the AI Act.
• Impersonate a real person or make the guest believe they are interacting with a human when they are interacting with the AI.
• Bypass the Service's security or human-oversight mechanisms.
• Exploit vulnerabilities related to age, disability or economic situation to influence a person's behaviour.
• Reuse, resell or integrate AI outputs in a high-risk AI system without MyKonci's prior written agreement.

Any breach of this policy constitutes a substantial violation of the T&Cs of sale that may lead to immediate suspension of the Service.

06Security and robustness

• The AI models used by MyKonci are provided by reputable vendors, subject to their own security and robustness requirements.
• System prompts are stored securely and their modification is logged.
• Regression tests are performed before any model or prompt update.
• Fallback mechanisms are in place in case of AI vendor unavailability.

07Training data and reuse

MyKonci commits not to use customer data to train or fine-tune external AI models, regardless of provider. Real conversations may be used to check the quality of the chatbot's replies (internal evaluation benchmarks). One customer's analyses are never used for another customer.

08AI incident reporting

Any incorrect, biased or inappropriate reply can be reported at [email protected]. Reports are handled within 48 business hours and are followed by analysis and, if necessary, a corrective update.

09Evolution

This policy will evolve as the AI Act progressively takes effect (August 2024 to August 2027) and as sector best practices mature. Any substantial change will be notified to customers.

10Contact

For any question about this AI policy, contact [email protected] or our data protection contact at [email protected].