AI Policy

March 31, 2026Read inFR EN ES DE IT PT

Article 1 — Purpose

The purpose of this Policy is to inform Users of the Hello Mira Platform about the use of artificial intelligence (AI) systems, in accordance with the transparency obligations set out in Regulation (EU) 2024/1689 on artificial intelligence (AI Act) and Regulation (EU) 2016/679 (GDPR).

Hello Mira is committed to a transparent, responsible and controlled use of artificial intelligence, in the service of its Users.

Article 2 — Hello Mira's role under the AI Act

Hello Mira acts primarily as a "deployer" within the meaning of the AI Act: Hello Mira uses AI systems developed by third parties (language model providers) in the context of its services.

Hello Mira also develops its own sovereign AI models, hosted on its own infrastructure (OVHcloud, EU). As these models take on a growing share of the functionalities, Hello Mira's role may evolve towards that of "provider" within the meaning of the AI Act, entailing additional obligations regarding technical documentation and risk management.

This Policy will be updated accordingly.

Article 3 — AI systems in use

3.1 Mira Brain — Proprietary AI engine

Technical nature: RAG (Retrieval-Augmented Generation) / CAG (Cache-Augmented Generation) system powered by Mira Knowledge, Hello Mira's proprietary knowledge base.

Operation: when the User asks a question, the system searches for relevant information in Mira Knowledge (verified field data, provided by Amigos, Mentors and the community), then uses a language model to generate a contextualised and personalised response.

Underlying models: Hello Mira relies on several language model providers (Mistral AI, Google Gemini, Anthropic Claude, OpenAI GPT) and is progressively developing its own sovereign models hosted within the European Union. The long-term objective is to reduce dependence on non-European providers.

Areas covered: visas and administration, health, culture, taxation, housing, security, travel planning.

AI Act classification: limited risk (AI system interacting with natural persons — Art. 50.1). Main obligations: transparency.

3.2 Specialised Mira Agents (Mira Chat)

Nature: conversational agents specialised by country and topic, powered by Mira Knowledge.

Function: provide contextualised information and advice to Nomads in specific areas (visas, health, culture, taxation, housing, security).

AI Act classification: limited risk — transparency obligations.

3.3 Identity verification (KYC)

Nature: automated passport analysis system, using optical recognition (MRZ extraction) and, where there is doubt about name matching, a vision model for investigation.

Function: verify the User's identity by comparing passport data against profile information.

Impact: this system makes decisions having an effect on the User's rights (identity validation or rejection, conditioning access to Mira Trips). As such, it is subject to Article 22 of the GDPR on automated decision-making.

Safeguards: a human review system is provided for all ambiguous cases. The User may contest any automated decision and obtain human intervention via customer support.

AI Act classification: limited risk — transparency obligations. The processing of biometric data (passport photo) requires enhanced vigilance.

3.4 Automated moderation (planned)

Nature: AI system for detecting content contrary to the Platform's rules (harassment, spam, illegal content, offensive content) in Mira Tribes and messages.

Status: planned, not yet deployed as of the publication date of this Policy.

Function: automated analysis of messages to detect CGU violations.

Impact: the system may trigger a precautionary temporary suspension of the account in the event of content manifestly contrary to the rules. Any permanent exclusion is the subject of a human decision.

AI Act classification: limited risk — transparency obligations.

3.5 Recommendation and personalisation (planned)

Nature: system for suggesting destinations, accommodation and experiences based on the User's profile and preferences.

Status: planned, not yet deployed.

AI Act classification: limited risk a priori — to be reassessed upon deployment.

Article 4 — Transparency obligations

4.1 Identification of AI systems

The User is always informed when interacting with an artificial intelligence system:

each Mira Agent is clearly identified as such in the Mira Chat interface (badge, explicit mention);
conversations involving an AI agent are flagged in the interface;
Mira Brain responses are identifiable as being AI-generated.

4.2 Limitations and warnings

The User is informed that:

AI system responses are automatically generated and may contain inaccuracies, incomplete or outdated information;
these responses do not in any way constitute professional advice (legal, medical, tax, financial, immigration);
Mira Brain is powered by Mira Knowledge (proprietary knowledge base) but does not guarantee the completeness or absolute currency of the information;
the User is encouraged to verify critical information with official sources (embassies, local authorities, healthcare professionals, tax advisors);
Hello Mira shall not be held liable for any decision made by the User on the basis of information provided by an AI system.

4.3 Reporting

The User may report any inappropriate, inaccurate or problematic response via the reporting mechanism integrated into the interface (report button available on each Mira Agent response).

Reports are reviewed by the Hello Mira team and contribute to the continuous improvement of the systems.

Article 5 — AI quality and governance

5.1 Mira Knowledge — Knowledge base

Mira Knowledge is a proprietary knowledge base, powered by:

field data collected from Amigos, Mentors and the Nomads community;
verified public sources (regulations, tourist information);
community experience feedback.

Hello Mira is committed to:

regularly updating the information contained in Mira Knowledge;
documenting the sources used;
flagging information whose currency cannot be guaranteed.

5.2 Human oversight

Hello Mira maintains human oversight over its AI systems:

automated decisions having an impact on the User's rights (KYC, moderation) are systematically accompanied by a human recourse mechanism;
the Hello Mira team conducts regular quality reviews of responses and conversations;
in the event that the Mira Agent is unable to respond satisfactorily, escalation to a human is possible.

5.3 Bias management

Hello Mira is committed to:

identifying and reducing potential biases in AI system responses;
ensuring equitable coverage of information regardless of the User's nationality, language or profile;
monitoring quality and satisfaction metrics to assess the relevance of responses.

Article 6 — Data and AI

6.1 Data processed by AI systems

During an interaction with a Mira Agent or Mira Brain, the following data may be used to generate the response:

the content of the question asked by the User;
the context of the conversation (previous messages in the session);
profile elements (travel preferences, nationality, destination consulted) if necessary for personalisation;
data from Mira Knowledge (knowledge base).

Sensitive data (passport scan, payment data, password) is never transmitted to conversational AI systems, except within the specific framework of the KYC process (Article 3.3).

6.2 Use for improvement

Conversations with AI systems may be used to improve service quality and enrich Mira Knowledge, under the conditions detailed in the Privacy Policy (Article 4).

The User has a right to object to this use (see Privacy Policy).

6.3 Transmission to AI providers

When Hello Mira uses third-party AI models (Gemini, Claude, GPT, Mistral), the content of requests is transmitted to the provider's servers for processing.

Hello Mira has verified with its providers that:

request data is not used for training the providers' models (except under specific contractual agreements providing otherwise, in which case the User is informed);
data is processed in accordance with SCCs (Standard Contractual Clauses) for transfers outside the EU.

Details of transfers are specified in the Privacy Policy.

Article 7 — Permissions system (Mira Agents)

Certain Mira Agents may request access to specific data to provide a personalised response (location, preferences, travel history).

The User:

is informed of each access request before consenting;
may grant or refuse each permission individually;
may revoke a permission at any time from their account settings;
may grant temporary permissions (limited duration).

Refusing a permission may limit the relevance of the Bot's response, without affecting access to the Platform's other functionalities.

Article 8 — Technological sovereignty

Hello Mira is engaged in a technological sovereignty approach aimed at:

developing its own proprietary AI models;
hosting its AI systems on European infrastructure (OVHcloud);
progressively reducing dependence on non-European AI providers;
ultimately ensuring that Users' data does not leave the European Union for AI processing.

This approach is progressive and will be reflected in updates to this Policy.

Article 9 — Changes and updates

This Policy will be updated in line with:

changes in the AI systems used by Hello Mira (deployment of automated moderation, the recommendation system, proprietary models);
changes in the regulatory framework (AI Act codes of practice, European Commission guidelines);
feedback from Users and supervisory authorities.

In the event of a substantial modification, Users will be informed by email and/or in-app notification.

Article 10 — Contact

For any questions regarding the use of artificial intelligence by Hello Mira:

Email: [email protected]
Postal address: CES Venture SAS — Euratechnologies, 165 avenue de Bretagne, 59000 Lille, France

To report a problem with an AI response: use the report button integrated into the Mira Chat interface.