IT And Research Safety Core

This core is mandatory for everyone working in the IDEAL Lab. It should take about 60-90 minutes, including the short scenarios.

Outcomes#

After completing the core, you can:

• protect accounts, secrets, private keys, and recovery codes;
• classify information before sharing it with cloud or AI tools;
• distinguish an AI model, gateway, agent harness, interface, and tool access;
• choose between your laptop, NAS, Blade, and Euler;
• inspect commands and diffs before accepting AI output;
• respond safely to a possible leak or access incident;
• ask for technical help without exposing sensitive information.

Modules#

Complete these in order:

1. Accounts and secrets
2. Data classification and AI tools
3. Laptop, NAS, Blade, and Euler
4. Incidents and getting help

Core Evidence#

The assessment asks you to review realistic scenarios rather than memorize commands. You will need to explain:

• why an SSH private key must never leave your computer;
• what to do when a token appears in Git history;
• whether a supplied research artifact may enter an AI tool;
• where source code, datasets, temporary files, and final results belong;
• why a plausible AI-generated command can still be unsafe.
• why a free or paid personal AI account does not approve project data.

Completion Check#

• I completed all four modules.
• I can identify the information owner when classification is unclear.
• I know the first containment step after a credential leak.
• I know that I remain responsible for commands run by an AI agent.
• I selected my capability tracks in START_HERE.md.