Security & Compliance in AI Apps Builder

AI and Data Access in AI Apps Builder

AI Apps Builder is an AI-based builder that uses a Large Language Model (LLM) provided through Anthropic Connect. The LLM is used to understand plain-English prompts and generate application code based on them.

• The AI does not analyze issues, comments, worklogs, or users

• The AI does not connect to Jira

• The AI does not call Jira APIs

• The AI does not receive real Jira data.

In AI Apps Builder, AI is used only to generate code:

• Generates an Atlassian Forge app

• Uses public Atlassian Forge documentation

• Describes which types of data are needed (for example: issue key, summary, due date)

• Builds the manifest.yml, UI components, and business logic.

Are Prompts Used to Train AI Models?

No. Neither Anthropic nor the AI Apps Builder team uses your prompts or generated outputs for model training. What you describe to generate an app does not feed back into improving the model.

This matters for teams handling sensitive internal processes, proprietary workflows, or confidential project structures. Describing your use case to build a custom app does not expose that information to future users of the system.

Important Note About the API Token

Where Is App Data Stored?

Apps generated with AI Apps Builder run on Atlassian Forge and use Forge-hosted storage — Storage API, Forge SQL, and similar mechanisms. This means:

• All data stays inside Atlassian's cloud infrastructure, not on any AI Apps Builder servers

• Each Jira site gets its own isolated storage space — one company's data is never mixed with another's

• Data is encrypted at rest using AES-256

• Data is backed up and managed entirely by Atlassian.

Data residency: If your Jira instance is pinned to a specific region (for example, EU or US), Forge-hosted app data follows that setting automatically. No additional configuration is required.

The egress exception: If an app makes calls to external services (such as third-party APIs or remote backends), data that leaves Forge falls outside Atlassian's infrastructure. In that case, you are responsible for the residency, security, and deletion of that external data — not Atlassian.

Forge security guarantees

Apps generated with AI Apps Builder are built and deployed on Atlassian Forge, which provides strong, built-in security guarantees by design. Forge enforces strict authentication and access control mechanisms. Only authorized users can access app data, and all access is governed by Jira’s standard permission model and the scopes explicitly granted during installation.

Forge apps run entirely inside Atlassian Cloud. This means Jira data does not leave Atlassian’s secure environment:

• No external servers are used

• No third-party backends are involved

• No data is sent outside the Atlassian infrastructure.

Forge apps inherit the same security framework that protects Jira. This includes:

• Platform-level security controls

• Ongoing security monitoring

• Compliance with Atlassian security policies and standards.

Scope Review Before Deployment

Apps created with AI Apps Builder use standard Forge scopes, selected based on the app’s functionality. Typical scopes may include:

You can review all scopes your app requests at two separate points before anything is installed:

First checkpoint — App Specification. Before the app is generated, AI Apps Builder creates an app specification that lists all requested scopes and permissions. You can review each scope, edit the specification, and confirm before the build begins. You can also narrow down scopes to match your organization's security requirements at this stage.

Second checkpoint — Deployment page. Before deployment, you get another opportunity to review all permissions. Nothing is installed without your explicit approval at each step.

If you use automatic deployment

If you use Manual deployment

Forge Shared Responsibility Model

As the AI Apps Builder generates a Forge app, you need to know that Forge uses a shared responsibility model:

Third-Party Integrations and Data Egress

AI Apps Builder supports connecting generated apps to external services such as Slack, Gmail, Google Drive, Google Calendar, HubSpot, and Notion. When an app includes third-party integrations, all external API calls are made from the Forge app running inside Atlassian's cloud — your credentials and data stay within Atlassian's security and permission model for as long as they remain inside Forge.

However, any data that leaves Forge and is sent to an external service falls outside Atlassian's infrastructure boundary. For such apps:

• You are responsible for the security, residency, and deletion of data processed by the external service

• You should review and document all external domains the app communicates with

• Atlassian's SOC 2 and ISO 27001 certifications do not extend to third-party services.

If your organization requires a "zero-egress" posture — where all processing and storage stays within Atlassian's cloud — you can build apps without third-party integrations. This model closely matches enterprise security expectations and can qualify for Atlassian's Runs on Atlassian program.

Code Transparency

You have full visibility into what AI Apps Builder generated:

• Download the generated code at any time from the deployment page as a ZIP file. You can review it, share it with your security team, or hand it to a developer.

• Review and edit the app specification before generation begins, including all modules, permissions, and scopes.

• Choose manual deployment and keep the full deployment process — including the API token — entirely under your control.