Skip to main content

How does Raison position itself with respect to the EU AI Act?

Understand how the EU AI Act classifies Raison's assistants and assessments, and how obligations split between Raison and institutions.

Written by Romain Morvant

The pedagogical assistants and agentic assessment activities offered by Raison fall within the high-risk category under Annex III, point 3(b) of the AI Act when deployed to evaluate student learning outcomes.

Provider obligations

Raison runs a compliance programme aligned with the obligations applicable to providers of high-risk systems:

  • Documented risk management system, kept up to date.

  • Data governance: quality, relevance, bias management on the corpora used to configure and evaluate assistants.

  • Technical documentation compliant with Annex IV (architecture, methods, limits, performance).

  • Automatic event logging. Technical logs are retained for 30 days (see How long is data retained?).

  • Transparency and information for end users (students, instructors).

  • Human oversight: interface design enabling arbitration and override by the instructor.

  • Robustness, accuracy and cybersecurity measured and documented.

  • CE marking and registration in the European database under Article 71.

Deployer obligations

Client deployers (institutions) remain responsible for their deployer obligations (monitoring operation, informing data subjects, retaining logs on the institution side). A dedicated AI Act addendum can be provided detailing the allocation of obligations between provider and deployer.

Related

Need help? Write to [email protected].

Did this answer your question?