Arbor AI FAQ's

This article addresses frequently asked questions about Arbor's AI functionalities, including data storage, security measures, compliance with regulations, user access control, and the operational capabilities of the AI. It emphasises that data processed by Arbor's AI is not stored or used for training models and that user initiation is required for all AI interactions.

What types of AI functionalities does Arbor offer?

Currently, the AI-powered Ask Arbor functionality is working in production

As part of Arbor Labs, we will be releasing test and in-development features to selected users (TBC).

Where and how is data stored and processed during AI operations?

• Arbor makes use of the Azure Open AI service. Any data processed by Arbor's AI functionality is not stored and isn't retained to train the AI model.
• The prompts (the inputs) and the completions (the outputs):
• Are not available/shared with other Microsoft Azure customers.
• Are not available/shared with Open AI (the makers of ChatGPT).
• Are not used to improve or train Open AI models.
• Are not used to improve any Microsoft or third-party products or services.
• Are not used to improve the Azure OpenAI models.
• The Azure Open AI Service is fully controlled by Microsoft; Microsoft hosts the Open AI model usings in Microsoft's Azure environment and the Service does NOT  interact with any services operated by OpenAI (e.g. ChatGPT, or the OpenAI API). 
• They are used exclusively by Arbor.

See the Azure OpenAI docs for more information.

Are there any security measures and protocols in place to protect data when using Arbor's AI functionality?

Arbor uses Azure's data centre, so no data leaves the UK. As such, they are simply another sub-processor.

Does Arbor's AI functionality comply with the data protection regulations and standards?

Yes, it is the same as any other sub-processor (see in the Where and how is data stored and processed during AI operations section of this article).

Is Arbor's functionality standalone?

Currently, the AI functionality is embedded within the Arbor MIS.

Please note: This may be changed in the future.

Can Arbor's AI functionality be turned off?

Depending on the user's access rights, they may or may not have access to our AI functionality (see T&C's, section 6). We make it very clear in our application where AI is used. Users can simply choose not to use it. When opting in, it is possible to control which users have access to the functionality. Further, the AI does not work independently to carry out actions or manipulate data on its behalf - actions are user-initiated or require human approval.

How is student data used with Ask Arbor?

When a request is made to Ask Arbour (our AI functionality), the AI model's response effectively instructs Arbor on how to complete the request.

To power Arbor's AI functionality, we use our own instances of the OpenAI models, which are hosted in the UK via Microsoft Azure. This means that any requests made to the models never leave the UK, and is processed entirely within Arbor's cloud infrastructure.

Any data processed by Arbor's AI functionality is not stored or retained to train the AI model.

What is the AI functionality in Arbor used for?

The AI functionality in Arbor is grouped into the following areas:

• Retrieving information (e.g. Ask Arbor assistant)
• Automating workflows (e.g. AI Absence Tool)
• Summarising information (e.g. Draft Me a Message)

How is access to Arbor's AI functionality controlled?

Initially, schools must choose to "switch on" Ask Arbor.

Once switched on, both the School: Ask Arbor permissions will be automatically added as Ad Hoc permissions to:

• Anyone who has the Head Teacher business role
• The person who has switched Ask Arbor on

Access to Ask Arbor is controlled and restricted in the same way as other features in Arbor. This means that access is determined on a per-user basis. Only the users granted permission by an administrator will have access to the functionality.

This also applies to access controlled student and staff data. If a user doesn't have access to certain fields, they will not be able to use Ask Arbor to view this data - the permissions to access data lies with the user making the request, not the AI agent.

All of the AI tooling in Arbor must be activated by a user. An AI tool cannot take action autonomously.

What does this mean?

Schools can choose to turn on Ask Arbor. If they don't choose to turn the functionality on, it will be disabled for all staff.

If a school's Arbor administrator chooses to turn on Ask Arbor, then they must permit individual staff members to use Ask Arbor. Ask Arbor isn't blanket-enabled for all staff. This means that access can be more tightly controlled by the school's Ask Arbor administrators.

For example: A school may only wish to enable  Ask Arbor for staff members who have undergone AI training and have signed the school's AI policy.

Can Arbor's AI functionality take action autonomously?

No. A user must always initiate any interactions with Arbor's AI functionality. All AI-generated content and/or actions must always be confirmed by a user.

For example, when creating an Ask Arbor 'Student Report':

• The user must click on 'Student Report' to initiate the report generation
• After the content is generated, the user is shown a preview of the report. They must then choose to share the report.

What kind of audit trails exist?

All Ask Arbor requests are logged by Arbor. The logs include:

• The user making the request
• Details of the request, including prompts
• Agent responses
• The "Tools invoked by the agent
• Any follow-up requests

Can the language for the Arbor AI be changed from American English to British English?

Currently, it is not possible to have the Arbor AI in British English. This is due to Arbor AI using the Azure OpenAI model, which is primarily trained on US data.