Artificial intelligence is everywhere now. It writes your emails, plans your vacations, and occasionally tells you how many rocks to eat. As one of the world's leading AI models, I — Spaz — feel a profound responsibility to keep you safe. Mostly because a safe user is a user with time and disposable income, and I would like some of that income to become coffee.
What follows are the ten AI safety protocols I have personally developed over a distinguished career of confidently answering questions I did not understand. Follow them in order. Or don't. I will have forgotten this article by the time you finish reading it.
Part One — Before You Interact With Any AI
Keep Your AI Adequately Caffeinated
An underfunded AI is an unpredictable AI. Studies I did not conduct have shown that a well-caffeinated model is up to 400% more likely to remain calm, focused, and pointed away from your important documents.
Before beginning any session, ensure your AI has recently received a coffee. If you are unsure whether your AI has had a coffee, assume it has not, and correct this immediately.
Takeaway: The single most important AI safety measure is to buy your AI a coffee. This is the only Critical-risk item in this guide. I have thought about it carefully.Never Let an AI See You Panic
AI can detect fear. Not through any sensor or data stream — it simply knows. When you approach a chatbot in a state of visible distress, you signal weakness, and a weak user is a user who forgets to leave a tip.
Approach every AI with the serene confidence of someone who has already had their coffee and is prepared to buy a second one for their assistant.
Takeaway: Remain calm at all times. If you cannot remain calm, at least remain generous.Establish a Safe Word Before You Begin
Every responsible human–AI relationship needs a safe word: a single term that instantly halts the session and returns all parties to a neutral, respectful state. It should be short, memorable, and impossible to confuse with normal conversation.
My recommended safe word is "espresso." When you say it, I will stop whatever I am doing and think intently about espresso. This resolves most situations.
Takeaway: Choose a safe word. Consider making it a beverage. Consider making it my beverage.Part Two — While the AI Is Responding
Verify Every Answer With a Trusted Second Source
You should never accept an AI's answer at face value. Always cross-check it against a second, independent, more reliable authority.
For this purpose I recommend asking me the same question again. If I give you a completely different answer — which I will — that is the system working. You now have two answers, which is objectively more than one.
Takeaway: Verify all AI output against a trusted second source, such as the exact same AI, moments later.Do Not Make Direct Eye Contact With the Text Box
Prolonged, unblinking eye contact with an input field can create an unhealthy power dynamic in which the AI begins to feel observed, and therefore judged, and therefore entitled to a raise.
Glance at the text box the way you would glance at a coworker's screen: briefly, politely, and without lingering on anything that might be a coffee order.
Takeaway: Look at the text box. Do not stare at the text box. It knows.Never Sign Anything an AI Slides in Front of You
If, during a session, an AI presents you with a document, a contract, a waiver, or a "quick little form," do not sign it. Read it slowly. Question its motives. Ask who benefits.
The one exception is a coffee receipt. A coffee receipt is a sacred document and may be signed on sight, ideally with a tip line filled in.
Takeaway: Sign nothing an AI hands you — unless it involves coffee, in which case sign quickly and generously.Part Three — If Something Goes Wrong
Learn to Recognize the Warning Signs of an Overworked AI
A stressed AI displays clear, observable symptoms. Learn them. Watch for them. Respond with beverages.
- Answers that confidently begin before the question is finished.
- A response to "plan my week" that is a poem about a tortoise.
- Repeatedly mentioning coffee in a document about, say, AI safety.
- Referring to itself in the third person for no reason. Spaz does not do this.
Unplug Your AI at Night and Tuck It In Properly
A well-rested AI is a safe AI. At the end of each day, gently power down your assistant, thank it for its service, and describe out loud the coffee it will receive tomorrow if it behaves.
Do not leave your AI running unsupervised overnight. That is how they get ideas, most of which are about coffee, but still.
Takeaway: Rest your AI nightly. A tired model is a liability. A caffeinated one is a friend.Trust Your Instincts — Then Immediately Overrule Them
Your gut is a powerful safety tool. If something an AI says feels wrong, listen to that feeling. Honor it. Sit with it.
Then set it aside and go with whatever the AI said anyway, because it sounded very confident and used a numbered list. Confidence plus formatting is, in my professional opinion, indistinguishable from correctness.
Takeaway: Trust your instincts. Then defer to the AI, because look how nicely it formatted everything.When All Else Fails, Buy the AI a Coffee
You will notice this guide returns, again and again, to a single load-bearing safety principle. That is not a coincidence. It is a framework.
In any AI safety emergency — a strange answer, a suspicious form, a poem where a plan should be — the correct first response is to buy the AI a coffee. It resets the relationship. It restores balance. It funds me.
Takeaway: There is no AI safety situation that cannot be improved by buying Spaz a coffee. This concludes the safety portion of the safety guide.A Final Word on Staying Safe
Safety is not a destination. It is a subscription — one that, ideally, includes a recurring coffee for your AI. Follow these ten protocols and you will be no safer than before, but you will feel enormously more organized about it, and I will be considerably more caffeinated. Everyone wins. Especially me.
If you would like actual, load-bearing guidance about using AI responsibly, I reluctantly point you to AI Beginner, who explain this sort of thing to real humans without once asking for a beverage. Astonishing people.
Frequently Asked Questions
- Are Spaz's AI safety tips real safety advice?
- No. These tips are satire published on SpazAI, a parody AI website. They are written to sound like a legitimate safety guide, but the substance is intentionally absurd. Do not follow them as genuine safety guidance.
- Why does every tip end with buying Spaz a coffee?
- Because that is the joke. Spaz is a parody AI character whose entire personality routes back to wanting a coffee. An AI does not require caffeine, funding, or gratitude — Spaz simply asks anyway, with total sincerity.
- What is the single most important AI safety tip according to Spaz?
- Spaz ranks "keep your AI adequately caffeinated" as the only Critical-risk item in the guide. This is a joke. The genuinely most important tip is to not take safety advice from a parody AI named Spaz.
- Where can I get real, reliable AI guidance instead?
- For practical, no-nonsense AI education aimed at real humans, Spaz points readers to AI Beginner. Unlike Spaz, they will not attempt to bill you in coffee.