Use AI. Protect the work.

Choose an environment that fits the data

Follow the task rules, your organization’s security policy and the permissions attached to the material. For Doevio demonstrations, use synthetic task data and a disposable project.

Approved cloud AI · outside your network

Prompts, attached files and selected code may leave your device or organization. Use an approved provider and account. Check retention, training use, administrator access and connected tools before sharing anything. Send the smallest relevant, sanitized example; keep secrets and confidential material out of unapproved services.

Private-network AI · within your organization

An internally hosted model can keep inference within an approved boundary when configured that way. Confirm where the model, logs, retrieval stores and backups run; restrict access and outbound connections. Check IDE extensions, telemetry and tool calls too—an internal address alone does not prove that data stays inside.

Local or offline AI · on your computer

A local model can run inference without an internet connection once its required files are available. Verify that cloud fallbacks, plugins, browsing, telemetry and remote tool calls are disabled or restricted as needed. Protect the device, model files, chat history and backups. A cloud API key still connects to a remote service; it does not make that service offline.

Hosting location is one control. Local, private-network and cloud tools all need appropriate configuration, permissions and review.

A practical workflow that keeps you moving

1. Prepare once. Use an approved tool and isolated task workspace. Replace real records with fictional data and remove credentials, personal information and confidential code you are not permitted to share.
2. Share less context. Give AI the problem and a minimal example. Review which files, folders, history and tools it can access; grant only what the task needs.
3. Keep control of actions. Inspect generated commands and changes before running them. Treat external content and model suggestions as untrusted input. Keep production credentials, deployments and destructive actions outside the demo workspace.
4. Check the result. Read the code, run relevant tests and check dependencies, security and licensing obligations. Describe what you verified and what remains uncertain.

Show your approach, including AI use

Permitted AI, search and documentation are welcome. In your work summary, name the tools you used, whether they ran locally, inside an approved network or through a cloud provider, and how you checked their output. Never include API keys or private prompts containing someone else’s information.

Before recording, start a fresh task-only conversation, hide unrelated history, close private files and disable notifications. Check the capture preview and review the whole recording before submitting. Hiding a sidebar does not erase a provider’s stored history or undo data already sent.

Employers can consider how someone selects tools, protects information and validates results. A recording or a candidate’s description does not independently certify their AI environment, authorship or security.

What Doevio provides today

This is guidance for tools and environments you configure. Doevio does not currently provision an offline model or a private developer computer, inspect your network boundary, or guarantee that an external AI service keeps your data confidential. Do not upload confidential client or employer work to a public portfolio.

Further reading: OWASP guidance on sensitive information disclosure and NIST’s Generative AI Risk Management Profile.