Test your confidential reporting channel yourself (not just theoretically) — is it actually anonymous, and did it respond the last time it was used?

Each team member: annual signature on disclosure of relationships affecting decisions.

In crisis: one person leads, all report to them, fast decisions — not "committee".

Identify 5 core values and verify your decisions actually reflect them.

Before ethically murky decision, evaluate on 4 axes: legal/ethical/values/reputation.

When an LLM helps you with content you publish or send, disclose it clearly.

Don't paste contracts, health data, or IP into a public model — may appear in later training.

Define 5 contexts where you'll NEVER use AI. Write them. Stick to them.

"AI text didn't copy anyone" is myth. Ensure you don't publish someone's idea/sentence unattributed.

Don't ask AI "write in author X's style". That = using their commercial style sans permission.

Write a personal list of situations where you exclude AI entirely — irreversible decisions needing purely human judgment.