Never paste secrets, credentials, or personal data
What must never enter a prompt, and the placeholder technique that lets you keep using AI safely.
Everything you paste into an AI prompt can leave your control. In IT that is a serious risk, because the raw material of the job is exactly the material you must protect.
The hard rule: never paste secrets, credentials, or personal data into a prompt. That includes passwords, API keys, connection strings, certificates, and tokens. It includes personal data such as employee records, email addresses in bulk, and anything covered by privacy rules. It includes internal details that would help an attacker, such as full network diagrams or the exact version of an unpatched system.
The technique that lets you keep using AI is placeholders. Replace the real value with a marker and wire the real value in yourself afterward. Instead of pasting a live connection string, write server equals PLACEHOLDER and key equals PLACEHOLDER. The model does not need the secret to write the logic; it only needs the shape.
The second technique is choosing the right tool for the data class. A personal chat account and an enterprise account that your organization has a contract with handle data very differently. For anything sensitive, prefer the approved, contractually covered tool, and know your data classification before you paste.
When in doubt, redact. A prompt that reads cleanly with fake values produces the same quality of script as one full of real secrets, without the exposure.
Take a command or config you would normally paste to an AI tool. Produce a redacted version where every secret, hostname, and identifier is a placeholder, keeping the structure intact.
Enjoying the free lessons? Get an email when we publish new courses and updates — no spam, unsubscribe anytime.
Discussion (0)
Ask a question or share what worked for you. Comments are reviewed before they appear.
No comments yet. Be the first to start the discussion!