AI at Wooclap involves the use of a new subcontractor — we have chosen OpenAI to ensure a high quality of service. However in the future we would like to use other sub-processors with strong security guarantees for more specific pedagogical purposes.
Wooclap Security team ran a Transfer Impact Analysis and we have implemented all necessary additional measures (technical, legal, and governmental) to ensure the security and compliance of data processing (you can also find this in the OpenAI Trust Center). You can find more information and guarantees regarding this processing in our Privacy Policy.
Wooclap is also using Amazon Textract, a document analysis service that detects and extracts printed text from images (for example, in our 'Label an image' question). Textract is an Amazon product integrated directly into our infrastructure, benefiting from all AWS security and privacy guarantees. The underlying machine learning model used by this tool has been trained on millions of documents.
Because of this, Wooclap wants to ensure that the AI technology aligns with our security standards. Therefore, users have the option to disable it through their account settings under the AI feature banner. In line with our principles, we have also disabled model training using Wooclap users' data (more information here).
The author of documents, content, questions, etc. remains the user's property (it is not Wooclap's property). Both ingested content (e.g. uploaded documents, audio, video) and AI-generated content remain the sole property of the user.
Wooclap users' data privacy is crucial for us. Currently, we do not use, nor do we intend to use, user-generated content or personal data to train AI models. Plus, we have contractual agreements with OpenAI that prohibit the use of Customer Data to train their models.
If tomorrow we consider using user-generated content to train AI models to improve quality, we will stick to the following essential principles:
- We will comply with European data privacy legislation (especially GDPR and EU AI Act).
- User consent for AI model training will be a prerequisite, and we will be fully transparent about the data used in each AI-related processing.
- We will be strict with our subcontractor, to ensure that the process does not use personal data for model training.
- Finally, model training will be an "opt-in" feature, meaning it will not be enabled by default.