FAQs
Access, eligibility, and rollout
Availability of AI coding tools varies by tool, organisation, and user type.
Refer to the AI coding tools page for the latest eligibility information.
Eligibility depends on the tool, your organisation, and your user type.
Refer to the AI coding tools page for the latest eligibility information.
Availability depends on the specific tool, your organisation, and your user type.
Refer to the AI coding tools page for the latest eligibility information.
Availability varies by tool and may depend on factors such as your organisation, user type, security requirements, and rollout status.
Refer to the AI coding tools page for the latest eligibility information.
No. Availability of agentic capabilities varies by tool.
Some advanced capabilities (such as CLI agents and background agents) may be limited to selected users during trials.
Refer to the AI Coding Tools Roadmap for current availability.
MCP is available for Claude Code (GovTech users only). Register at go.gov.sg/mcp-gateway.
For the latest supported MCPs and setup details, see #claude-code-support on Slack.
MCP remains disabled for other tools.
GovTech ARs must be sponsored by a Team Lead, provisioned with a Platform AI key, complete the onboarding form, and follow the onboarding workflow described in Onboard GovTech ARs to Claude Code.
For Augmented Resources (ARs), access to agentic AI tools may take up to 2 business days after form submission. Please wait at least 2 business days before raising a support ticket if you have not received access.
Some AI coding tools are introduced to selected user groups first to validate safeguards, gather feedback, and refine operational processes before broader adoption.
Rollout approaches may differ across tools depending on governance requirements, security considerations, and operational readiness.
Security and Sensitive Classification
AI coding tools support different classification and sensitivity levels.
Refer to Security and Sensitive Classification for the latest supported classifications, sensitivities, and approved tools.
The classification of a system is based on the sensitivity of the overall system and its information assets.
The classification of source code is determined by the logic and information contained in the code itself. As a result, source code classification is not always the same as system classification.
Refer to the Code Classification Overview for guidance on classifying source code.
Source code classification should be determined in accordance with your agency’s governance process.
If you are unsure, consult your project lead, security team, or the appropriate governance authority within your agency.
Refer to the Code Classification Overview for guidance.
Not necessarily.
Different AI coding tools support different classification and sensitivity levels.
Refer to Security and Sensitive Classification to determine whether your classification is supported by the tool you intend to use.
Yes, but additional system-level approval is required.
If your system is classified as Critical Information Infrastructure (CII) or Significant Information Infrastructure (SII), you must obtain the required approvals before using agentic AI tools, even if the tool itself is approved for your code classification.
Refer to Additional approvals for details.
Tool approval and system approval are separate requirements.
An AI coding tool may support the classification and sensitivity level of your source code, but additional approvals may still be required for systems classified as CII or SII.
Refer to Additional approvals for details.
GitHub Copilot has access to the tabs currently open in your workspace. If you don’t open certain folders, those folders aren’t accessed by Copilot as part of the current workspace. However, if you use the @workspace command to allow Copilot to traverse directories, it might access other folders, including sensitive ones.
Code security and privacy
No. GitHub Copilot for Business does not use your code or prompts to train the service.
For more details, refer to the GitHub Copilot Trust Center.
Yes, when using Copilot for Business, some code snippets are sent to the cloud to generate responses, and these snippets are promptly deleted after the response is generated. However, GovTech does not recommend opening workspaces containing RESTRICTED code or secret tokens while using Copilot, as the extension may access open files and tabs to provide relevant suggestions.
Copilot has a security filter that scans your code for API keys and other sensitive data before sending it to the cloud. However, it’s crucial to avoid including API keys directly in your code and to use environment variables or approved secret-management solutions for storing sensitive information.
Protecting your code and data remains your responsibility.
No. GitHub Copilot for Business does not send any data back to OpenAI. It does not use your data for training purposes, and your data remains secure and private within the GitHub infrastructure provided by Azure.
Telemetry helps evaluate the effectiveness, adoption, and operational performance of AI coding tools.
Examples include:
- Usage patterns
- Feature adoption
- Acceptance rates
- Operational metrics
Telemetry data supports governance, service improvement, and platform planning.
Usage and configuration
Yes. However, you must implement your own security guardrails and obtain the required policy clearances.
For details, see Use other AI coding tools.
We recommend using licences provisioned through approved organisational channels.
Using a personal licence may prevent telemetry and governance controls from being applied and may not satisfy organisational security requirements.
No. Opting out of agentic AI does not allow you to disable Cloudflare WARP.
The Cloudflare WARP “always-on” policy is being progressively rolled out to all SEED devices, regardless of AI tool usage. Once the policy is applied to your device, you cannot disable WARP.
Cloudflare WARP is a mandatory security baseline that ensures secure, encrypted access to internal and external resources.
User provisioning in GitHub Enterprise can take some time to complete. To ensure your access is set up properly, please allow up to 40 minutes after receiving the welcome email before trying to sign in.
After 40 minutes, if you are still unable to sign in, please contact SHIP-HATS.
To resolve this issue, try enabling Experimental System Certificates V2 in VS Code.
To enable Experimental System Certificates V2 in VS Code:
- Open VS Code settings and search for System Certificates V2.
- Select Http > Experimental: System Certificates V2.
- Restart VS Code.
Product and feature support
Eclipse is not currently supported because of limited vendor support and lower adoption compared to other IDEs.
As an alternative, consider IntelliJ IDEA or IntelliJ Community Edition with Eclipse key bindings.
Yes. However, you should ensure the classification and sensitivity level of your workload are supported for your intended use case.
Refer to Security and Sensitive Classification for guidance.
Copilot does not scan your project folder automatically by default.
To include project context, explicitly reference files in the chat or use features such as @workspace, where supported.
For more details, refer to the GitHub Copilot documentation.
Follow the instructions in the GitHub Copilot documentation to install, sign in to, and authenticate the GitHub Copilot extension.
You can continue to use your GitLab credentials to manage repositories independently.
Yes. You can enable or disable the GitHub Copilot extension as needed.
Refer to the GitHub documentation for the latest instructions.
Currently, the @workspace command is not available in IntelliJ.
Refer to the GitHub Copilot documentation for the latest feature availability.
Refer to the GitHub Copilot documentation for installation, authentication, and setup instructions for supported JetBrains IDEs.
