Please explain how data is handled regarding AI features

Is information or data entered when using Teachme Biz's AI features used to retrain the AI model?

No, your data is not used to retrain the AI model.
We have implemented a system that prevents data from leaving Teachme Biz.

Related: Is information entered or files uploaded when using AI features used to retrain the model?
https://help.teachme.jp/hc/en-us/articles/34973469783961

How are uploaded PDFs, videos, and text stored?

Except for certain features (※), uploaded data itself is not stored.
※ For the feature that creates manuals from PDFs, the media library feature, and the studio feature, uploaded data is stored in S3 (Amazon Simple Storage Service) for the purpose of customers using it to create and edit manuals.

Videos and text registered in steps are stored in AWS (Amazon Web Services) for the purpose of being used as manuals within the customer's environment.

Intermediate outputs (such as OCR results and responses from LLMs) are stored in AWS for the purpose of providing the AI chat feature and for identifying issues and improving accuracy.

Please note that as the service's features expand, the method and duration of data storage may change.
If there are any significant changes, we will notify you separately.

Is there any risk of data being mixed with that of other companies?

In Teachme Biz, data within Teachme Biz is separated on a per-company basis, so it will not be mixed with data from other companies.

Related: Are the dedicated and shared portions clearly separated for each company using the service?
https://help.teachme.jp/hc/en-us/articles/7437500113817

Who owns the copyright to generated manuals?

The copyright to generated manuals belongs to the customer.
Teachme Biz's AI features function as a tool to assist in generating content based on the customer's input.

For details, please refer to Article 18: Protection of Contractor Content in the Teachme Biz Terms of Service.

▼ Where can I find the Teachme Biz Terms of Service?
https://help.teachme.jp/hc/en-us/articles/900005509406

How is the quality of AI-generated results ensured?

We verify whether the intended generation results are achieved with a certain level of probability or accuracy, tailored to the type of AI feature.

However, due to the nature of AI, it is not possible to fully control the generated results.
The system is designed so that generated results can be reviewed and edited before publication.
Please perform final quality control on your end.

How are cases of inappropriate AI-generated content handled?

In the unlikely event of inappropriate generated content, it can be corrected or deleted immediately.
Additionally, if you report it to our support desk, we will use the information to improve our system.

What generative AI models are being used?

We use Anthropic's Claude model in multiple AI-related features, and depending on the feature, we also use models from Amazon, Google's Gemini model, Whisper, and others in combination.

We review the AI services and models we use in line with advances in AI technology, and continuously improve our system configuration to always provide customers with appropriate performance.
The models we select are evaluated comprehensively based on performance as well as their track record in enterprise use and reliability from a security standpoint.

Please note that inputted data is not used to retrain the model.