🔐 Rule 1 · The "Don't Share" List

Some things never go in an AI chat.

AI chats can be saved, logged, used to train future models, or breached. Treat any AI chat the way you'd treat posting on Facebook — assume someone might see it. Then live by this short list.

🔐
Social Security numbers. Bank info. Credit cards. Passwords.
Never. Not for any reason. Not even to "test" what AI would say.
🏣
Your full home address. Kids' full names. Kids' schools.
If you need help with a school issue, say "my child's school" not the school name. Say "my neighborhood" not the street.
🏪
Deployment dates. Specific unit names. Exact orders dates.
Use a window ("next quarter," "late summer") instead of a date. This is the OPSEC rule that protects your spouse and your unit.
🏥
Detailed medical records. Diagnoses tied to your name.
AI can help you understand a diagnosis or prep for an appointment — just describe symptoms or general situation, not your full medical chart.
👨‍⚗️
Other people's private info without their consent.
Your sister's drama. Your boss's medical issue. Your kid's IEP details with the kid's full name. Anonymize before you paste.
The one-line rule

If you would not put it on a bulletin board in your church or your unit, do not put it in an AI chat.

🔎 Rule 2 · The Smell Test

AI can be confidently wrong.

Claude is good. It is not perfect. Read every answer with these four checks before you act on it.

1
Does it cite something specific?

If it quotes a law, a date, a court case — verify it. AI sometimes invents real-sounding sources that do not exist.

2
Does it sound too sure?

If it gives you one answer with zero hedging on something complicated — medical, legal, financial — that is a flag. Push back. Ask for the counter.

3
Does it match your gut?

If it tells you something that contradicts what you already know to be true — trust yourself first. Ask Claude how it knows.

4
Would a human expert agree?

For medical, legal, IEP, or financial decisions — use Claude to prep for the conversation with the expert, not to replace them.

When to push back on Claude

Tell it: "Are you sure? Show me how you know that." Or: "What is the strongest argument against what you just said?" Claude will tell you when it is not sure. You just have to ask.

✋ Rule 3 · When to walk away

Some tasks are not AI tasks.

AI is great for writing, brainstorming, summarizing, drafting, translating. It is not where you go for these:

🧡
A real crisis or emergency.
If you or someone you love is in danger, hurting themselves, or thinking about it — call 988 (Suicide & Crisis Lifeline) or 911. Military families: Military Crisis Line is also 988, press 1.
⚖️
Final medical decisions.
Use Claude to understand what your doctor said, prep questions, or read a lab result — but not to make the call.
⚖️
Final legal decisions.
Use Claude to understand a contract or get organized for a hearing — not to sign anything binding without a lawyer.
💰
Specific stock trades, crypto bets, or "guaranteed" anything.
AI does not know the future. Anyone telling you otherwise is selling you something.
👋🏾
Replacing the human in your corner.
Therapist. Pastor. Mentor. Best friend. AI helps. It does not replace the people who know you.
🎖 Rule 4 · Military Family OPSEC

For mil-spouses — the rule that protects everyone.

OPSEC is not paranoia. It is operational security — the practice of not giving away information that could put your spouse, their unit, or your family at risk. Here is how it applies to AI.

📍
Use windows, not exact dates.
Instead of "deployment August 12-November 4," say "deployment late summer through fall." Claude can still help with logistics; the window keeps you covered.
🏪
Use general unit names, not specific.
"Army infantry unit" is fine. "1-509 PIR Bravo Company" is not.
✈️
Do not name training exercises, downrange locations, or destinations.
"Overseas training" or "downrange" is enough for the prompt to work.
👨‍👨‍👧‍👦
Talk to your spouse before you post or share any AI-generated content publicly.
Especially anything that mentions deployments, training, or family adjustments. Their command may have specific rules.
The OPSEC golden rule

Would you say this to a stranger at the gym? If not, do not say it to Claude. AI is helpful — but it is still a stranger.

Now you know the rules. Go play.

The AI Family Room is built around these guardrails on purpose. Every mission keeps you safe by default. Pick one and earn your first badge.

Show me the missions →