Where does the responses generated by the AI comes from?
The feature comes from the training data 360Learning provides to Microsoft Azure Open AI exclusively for the creation of customized models and the provision of the services.
For more information, visit the Data, privacy, and security for Azure OpenAI Service page ↗.
Does the AI collect the data I share?
Yes, only for the provisions of the Services and for the duration of the contract. Some data, such as prompts and completions, is stored for a maximum of 30 days.
For more information, visit the Data, privacy, and security for Azure OpenAI Service page ↗.
Where is stored the data collected by the AI?
The data is stored in the Microsoft Azure Data Centers located in France.
For what other purposes will be used the data I share with the AI?
Azure OpenAI supports 360Learning AI-assisted authoring features.
As of August 1st, 2024, this corresponds to:
- AI-suggested questions, also known as QGen
- AI-powered course creation from a PDF document
- Suggested skills and courses (for activated platforms)
Will the collected data be shared with third parties?
No. the data is exclusively used for the performance of the services provided by 360Learning. The customer data is not used to train, retrain or improve the models in the Azure OpenAI Service.
Do you have the legal right to use AI?
Yes. As per our contract, 360Learning has the right to make updates, bug fixes, enhancements, new releases, new versions, and other improvements to the Platform.
Furthermore, we are using Microsoft Azure Open AI technology, Microsoft Azure is already one of our sub-processors and the use of its Open AI services does not alter the contract provisions.
Customers can choose freely to use or not use this feature.
Does 360Learning plan to replace jobs with AI?
No.
Is the usage of AI by 360L GDPR-compliant?
Yes. We are using Microsoft Azure Open AI technology, Microsoft Azure is already one of our sub-processors and it's GDPR compliant. The use of its Open AI services does not alter the contract and Data Protection Agreement (DPA) provisions we have in place.
How do you ensure the content is correct and has no mistake?
We have opted for the implementation of "high" filters and control functionalities by Microsoft Azure Open AI. Standard filters (high & medium) lead to too many “false positives”.
Reference data also helps the model to avoid generating responses based on obsolete or incorrect information.
We encourage customers to check generated content before publication.
By way of information, we are currently developing a feature on the platform offering "automatic" verification of content conformity, in addition to Open AI filtering. This check complements the filtering by allowing personalization by reference to internal policies.
What AI technology does 360Learning use?
Azure OpenAI Service gives customers advanced language AI with OpenAI GPT-4, GPT-3, Codex, and DALL-E models with the security and enterprise promise of Azure. Azure OpenAI co-develops the APIs with OpenAI, ensuring compatibility and a smooth transition from one to the other.
With Azure OpenAI, customers get the security capabilities of Microsoft Azure while running the same models as OpenAI. Azure OpenAI offers private networking, regional availability, and responsible AI content filtering.
More information available at: https://learn.microsoft.com/en-us/azure/cognitive-services/openai/overview#key-concepts.
Additionally, for "Skills" services customers, Transformers custom models are also used. These models are based on each customer data and combined with Azure OpenAI to provide the best personalized experience to end-users.
What data does the Azure OpenAI Service process?
- Prompts and completions
- Training & validation data
- Results data from training process
More information available at: https://learn.microsoft.com/en-us/legal/cognitive-services/openai/data-privacy?context=%2Fazure%2Fcognitive-services%2Fopenai%2Fcontext%2Fcontext.
Is customer data processed by Azure OpenAI sent to OpenAI?
No. Microsoft hosts the OpenAI models within our Azure infrastructure, and all customer data sent to Azure OpenAI remains within the Azure OpenAI service.
Is customer data used to train the OpenAI models?
No. The model is the result of the training on generic data.
It is configured by establishing specific prompts by the client. This configuration allows the model to generate courses that consistently follow the pre-defined instructions (systematic prompts) generating course content by reference to the submitted documentation.
The training data is not composed of prompts, documents (“inputs) and/or generated training (“Outputs”). In other words, those elements are not used to train, build or enhance the sample data of the Microsoft Azure Open AI Service or 360Learning Model.
Not accessible to other users.
Not accessible by OpenAI.
Not used to improve OpenAI models.
Not used to improve Microsoft or third-party products or services.
The data flows are only inbound. No data relying on the Chat GPT knowledge is shared.
How is data processed to build recommendations?
The model used for generating personalized recommendations is LightFM, an open-source model.
To comply with data privacy, the model is trained with anonymized features. These include:
- User ID: A unique identifier for each user on the platform.
- Item ID: A unique identifier for each course or learning path.
- Item Language: The language of the course or path.
- Group ID: Unique identifier of groups the user belongs to.
- Tags ID: Unique identifiers representing tags associated with courses or paths.
Each platform has its own dedicated model, trained on its own proprietary data. There is no data mixing across different platforms.
How do you manage the segregation/privacy of the data used to train the models?
Your fine-tuned Azure OpenAI model is available exclusively for your use.
See more at: https://learn.microsoft.com/en-us/legal/cognitive-services/openai/data-privacy?context=%2Fazure%2Fcognitive-services%2Fopenai%2Fcontext%2Fcontext.
How do you manage access to AI functionality within my organization? How do you manage the separation/confidentiality of data used between users?
By implementing role matrices within the platform, it is possible to reserve AI functionalities for certain types of users.
For example, the creation and definition of customized prompts can only be available to certain categories of users, such as administrators or content creators.
Groups can also be used to manage access to generated training courses, thus preserving the confidentiality of generated content.
What's more, AI functionalities respect the same visibility rules as all functionalities on the platform. Data sent to AI via a PDF is only visible to the content creator, the learners to whom this module is made visible, and the advanced roles (admin, coach) who have visibility over this module.
An option allows the content creator to attach the PDF used to generate the module, at the end of the module, in the "Summary" section, enabling learners with access to this module to download the document. In any case, the author of the module can decide to manually delete the PDF attached to the module at any time.
Is the ChatGPT Azure Open AI engine connected to the public version?
The model is not connected to the public version. It uses the paid version of Azure Open AI Service (Chat GPT 4o) for enterprises, and we have subscribed to the Azure Open AI offer for private networking.
Do I own the data generated from its data?
Yes. In accordance with the contract concluded with 360Learning, you retain ownership of the rights to the content and results generated, a principle that may be reiterated in an addendum governing the use of AI.
How will this functionality fit into the platform? What elements change or are enriched?
360Learning offers the following feature options using generative AI technology:
- AI-suggested surveys (also known as QGen) and based on content uploaded by content-creating users.
- AI-based module creation from PDF documents
- Application of customized prompts with instructions created by administrators.
- For customers using "Skills" Services: features to generate skills ontology and extract skills from customer courses, job descriptions and user experiences.
Those features are offered as part of new product plans.
Which kind of content does your AI manipulate (image, text, array, drawings…)?
AI functionalities for creating training from uploaded content only concern text-to-text generation and speech-to-text generation.
Our teams are exploring the possibility of using other AI content such as images and videos in the future.
The platform chat/forum does not use AI.