Artificial intelligence at Dastra
Principles for transparent and responsible AI
Last updated: October 2026
Introduction and definitions
Dastra uses artificial intelligence in its internal operations. AI is also used in certain features of its SaaS platform.
This statement sets out the principles and safeguards that guide how AI is made available and used within Dastra and on the SaaS platform it provides to its customers.
For the purposes of this statement:
“AI”, “Artificial intelligence” and “AI systems” mean software or automated systems used by Dastra employees in the course of their work or made available to customers on the Dastra platform. These systems can, in particular, generate content, suggestions, recommendations, summaries, analyses or responses, or carry out automated actions or tasks.
“AI output” means any content, prediction, suggestion, recommendation, summary, analysis or decision-support material produced by an artificial intelligence system.
1. Use of AI within Dastra
Dastra employees use artificial intelligence to support their work, including finding and summarising information, drafting documents, developing software and improving certain internal processes. AI tools complement the teams’ working methods and expertise without replacing them.
1.1. Confidentiality of information
Employees are not authorised to share customer data with AI tools used for internal purposes. This rule applies in particular to documents, personal data and other confidential material entrusted to Dastra by its customers, as well as third-party intellectual property rights.
The same care applies to internal personal data and Dastra’s sensitive information.
1.2. Approved tools and good practices
Employees use the AI tools approved and listed by Dastra, in accordance with internal policies and applicable procedures. When using these tools, employees take their characteristics into account, particularly their terms for confidentiality and data processing.
1.3. Human review and validation
AI remains an assistive tool. Content, documents, analyses, code and other work produced with its assistance are always reviewed and validated by a person before use or distribution.
This review assesses the accuracy, relevance and quality of the output.
When developing the platform or carrying out AI-assisted work that may affect the services offered to customers, Dastra applies its usual security and quality requirements, including code review, functional testing, verification of results, data protection, access control and vulnerability prevention.
AI-assisted work is subject to the same verification and validation procedures as work produced without AI before Dastra integrates it into the platform or deploys it to production.
1.4. Employee training and awareness
Dastra trains its employees and raises their awareness of the capabilities, limitations and risks associated with using artificial intelligence systems.
This approach includes training and awareness initiatives on good practices, particularly in confidentiality, data protection, cybersecurity, respect for third-party rights and the reliability of generated outputs.
The aim is to foster responsible internal use of AI based on critical thinking, professional judgement and verification of outputs before they are used or distributed, or both.
2. AI features available on Dastra’s SaaS platform
The Dastra platform includes artificial intelligence features to help business customers with data protection, compliance and governance activities.
2.1. Main use cases
These features include:
- Content generation, including drafting, rephrasing, summarising and analysing documents or information related to customers’ compliance activities.
- Conversational assistance, including context-aware responses, information retrieval and help with certain tasks.
- Automation through AI assistants, which automatically perform certain actions on the platform according to the settings and permissions defined by customers.
2.2. Different levels of autonomy
These uses have different levels of autonomy. Some AI features provide content or suggestions for users to review and validate. Others can carry out automated actions on the platform within the permissions and settings configured by customers.
The AI features available on the Dastra platform are designed to support customers in their professional activities. Customers remain responsible for putting in place control and validation procedures suited to their uses, circumstances and associated risks.
3. Transparency and information
3.1. Informing users
Users must be able to recognise AI-powered features on the Dastra platform and understand their purposes.
When users use an AI system on Dastra, they are immediately informed that they are interacting with AI and can quickly identify what the system can be used for. Users will not be led to believe that an AI system is a human interlocutor. The information provided by Dastra is intended to help users understand how the AI features work, what they are for and what their limitations are, so they can use them in an informed and responsible way.
3.2. Identifying AI-generated outputs
AI-generated outputs are presented in a way that allows users to quickly identify their origin and distinguish them from content created entirely by people. On the Dastra platform, this may take the form of an explanatory notice or an “AI” tag associated with the relevant content.
3.4. Informing third parties
When a customer deploys AI systems that interact with third parties, including data subjects under Regulation (EU) 2016/679 on the protection of personal data (GDPR), the customer is responsible for providing them with the transparency information required by applicable regulations.
4. Personal data and confidentiality
4.1. Data processed by AI features
The artificial intelligence features available on the Dastra platform use the data necessary to respond to user requests, generate content or perform the requested operations. Dastra takes care to limit the information sent to artificial intelligence models to data that is necessary and relevant to the AI feature being used and the need expressed by the user.
Personal data processed in connection with Dastra’s artificial intelligence features is subject to the contractual terms that apply between Dastra and its customers, including those relating to the GDPR.
4.2. No use of customer data for model training
Customer data processed in connection with the use of artificial intelligence services provided by Dastra is not used to train, retrain or fine-tune artificial intelligence models.
4.3. AI model providers
Dastra pays particular attention to its choice of artificial intelligence technology providers, taking into account their practices in areas including personal data protection, confidentiality, information security and transparency.
Providers are selected based on these criteria, their commitments and the safeguards they offer, in accordance with applicable regulatory requirements.
Using AI features may involve sending AI models the instructions and contextual information needed to perform the requested operations. The nature and scope of this information vary according to the features used.
5. Risk management and security of AI systems
Information security and data protection are integral to Dastra’s approach, which draws in particular on its ISO/IEC 27001 certification for information security management and ISO/IEC 27701 certification for privacy information management.
These principles are taken into account in the design, development and continued evolution of AI features, with particular attention to data confidentiality, access management and control of the risks associated with their use.
6. Human oversight and control of AI systems
Human oversight is tailored to the nature of the artificial intelligence features and their level of autonomy. It is intended to enable users and customers to control the use of AI, assess the outputs and oversee the operations carried out.
6.1. Settings, roles and permissions
Customers can control access to AI features through roles and permissions and define the permissions associated with certain use cases. Customers may also request that AI features be disabled for their organisation at any time.
These options allow customers to decide what role artificial intelligence should play in their processes, taking into account their internal rules, the users involved and the risks they wish to manage.
6.2. Human control and oversight
Depending on how they operate by default and how the customer configures them, some artificial intelligence features require prior human validation. Others, particularly AI agents, can perform operations autonomously without requiring validation of every action.
Human oversight must therefore be tailored to the level of autonomy of the features and the settings defined by the customer. Control relies in particular on managing access rights and permissions, configuring how the features are used, understanding the AI features and being able to intervene when necessary.
7. Limitations of outputs
Outputs produced by artificial intelligence systems may be inaccurate or incomplete and may contain errors, omissions or approximations. These outputs support customers’ decision-making and cannot, on their own, establish that a customer complies with its legal and regulatory obligations.
Users are therefore encouraged to verify outputs before incorporating them into their documentation, risk assessments or operational processes. This verification must take into account the context, the available information and applicable legal and regulatory requirements, drawing on appropriate expertise where necessary.
8. Fairness, non-discrimination and fundamental rights
Artificial intelligence systems can reproduce biases and generate inaccurate, misleading, discriminatory or unfair outputs.
Dastra takes these risks into account from the design and development stages of its AI features and seeks to apply good practices and security measures to limit them. Dastra pays particular attention to privacy, the principle of non-discrimination, fundamental rights and third-party intellectual property rights.
AI outputs do not replace users’ judgement. Users are responsible for reviewing them with discernment and professionalism to prevent the risk of unfair or inappropriate decisions.
9. AI governance
AI governance at Dastra is based on an understanding of the uses, capabilities, limitations and risks associated with these technologies, along with a clear definition of each party’s responsibilities. The principles of transparency, proportionality, risk management and human oversight guide Dastra’s internal practices as well as the design, development and continued evolution of the AI features offered to customers.
10. Monitoring, reporting and continuous improvement
10.1. Monitoring and correcting anomalies
Dastra regularly tests its artificial intelligence systems to detect anomalies, identify risks and correct malfunctions. These actions contribute to the continuous improvement of the security, reliability and performance of the AI features offered by Dastra.
10.2. Incident reporting
Users are encouraged to report any incorrect output, unexpected behaviour or incident encountered while using AI features to Dastra Support. These reports allow Dastra to analyse the issues users encounter and continuously improve its services.
11. Updates and scope of this statement
This statement may be updated to reflect changes in services, internal uses, the technologies used or the applicable regulatory framework. The latest version of this statement is published on Dastra’s website, together with the date it was last updated. Where appropriate, changes may be communicated separately to customers.
This statement is for information purposes only. It is not a contractual document and does not amend or supplement the contractual commitments between Dastra and its customers.