Start with a private workspace, minimal data and the smallest number of services. Do not make public deployment the default reward for completing lessons.
First-session launch checklist
- Named mentor and adult escalation contacts confirmed.
- Trial or project agreement accepted.
- Exact tools checked against the learner’s age and account type.
- Required permissions completed without false ages or shared logins.
- Parent-visible contact and session access tested.
- Recording and transcription settings checked.
- Software installation tested on the actual learner’s device.
- Synthetic, non-identifying or appropriately cleared project data selected.
- No employer material, classmates’ information or personal images included.
- Workspace visibility checked; public sharing disabled unless separately approved.
- Files backed up in an agreed location.
- Spending settings and shutdown responsibility documented.
- Baseline activity and contribution log ready.
If using an API, keep secret keys out of browser code, mobile applications, source control and logs. OpenAI’s official API guidance explicitly identifies these as places secrets should not be exposed. Ask the mentor to demonstrate the actual arrangement rather than merely state that keys are “secure.” (developers.openai.com)
Verify whether a budget actually stops usage. OpenAI’s current documentation distinguishes spend alerts from enforceable hard limits; hard-limit enforcement is not instantaneous and spending can slightly exceed the configured amount. Use the available enforcement setting, monitor usage and maintain a way to disable the project. Do not rely on an alert as a guaranteed cap. (developers.openai.com)
Sources cited on this page
- API keys and service accounts | OpenAI API ↗developers.openai.com
- Spend limits | OpenAI API ↗developers.openai.com