The Real Reason AI Lies To You
How I almost sent my daughter’s landlord a fake law—and what it taught me about leadership.
💌 If this made you think, forward it to a founder who still thinks “AI lies.” Because maybe, like me, they just need better instructions.
The Day My Chat Friend Lied to Me
This summer, I almost gave my daughter very wrong legal advice.
She’s in her second year of university. She’s newly independent and the proud renter of her first “single-lady” apartment (cat and all). She pays rent on time and keeps a spreadsheet for her budget. She is, in every way, a responsible adult.
Her landlord, however, is not.
One night, walking home from work, she lost her mailbox keys. Not a big deal, right? Until her landlord demanded payment equivalent to her entire monthly paycheck for a replacement.
The same paycheck that was, ironically, stuck in her mailbox.
So she did what any smart university daughter of mine would do—she sicced her lawyer mom on him.
Her lawyer mom (that’s me) sat down to write a firm, righteous letter reminding him of his legal obligations and moral deficiencies. But before hitting “send,” I did what every modern woman does when she’s too tired to research case law at 10 p.m.
I asked ChatGPT to draft it.
That’s when AI scared the poop out of me.
I asked the bot to cite the law on replacement keys in Ontario. It responded confidently, like a first-year associate trying to impress a senior partner:
“Under section 23(4) of the Ontario Residential Tenancies Act, a landlord may not charge more than the cost of replacement.”
Perfect, right? Lawyering is so easy nowadays, right?
Except…I knew better. So I pulled up my big lawyer pants and did my own AI-free research.
You may be surprised to hear what I learned:
→ a) Section 23(4) doesn’t exist, and
→ b) The only thing the law says about keys is that landlords can’t change the locks without giving tenants a replacement key.
The “source” it cited? A random PDF from a real-estate conference in 2011. Not a statute. Not even close.
I pushed back. I politely told ChatGPT it was wrong. But it doubled down.
Rude!
Such a liar!
To add insult to injury, if I hadn’t double-checked, my chatbot would’ve had me send a legal letter full of imaginary law… on behalf of my own daughter. 😳
TBH, this scared me so much my first reaction was to tell the world to stop using AI forever.
But after I calmed down, I realized two things I want every business owner to know:
1️⃣ AI hallucinates.
2️⃣ Humans need to use AI better.
AI Didn’t Fail Me. I Was Just a Bad Boss
Why does this matter for business owners? Because AI didn’t lie to me. It did exactly what I told it to do. It found words that sounded right and delivered them with confidence.
AI doesn’t know truth; it predicts patterns. It’s a mirror that reflects our instructions.
When people say “AI hallucinates,” what they really mean is: we gave it too much room to guess.
And that’s exactly what I’d done. TBH, my prompt was lazy.
I Needed to Train Myself to Talk to the Bot
The next morning, I started over with a real prompt.
And just like that, my “hallucinating” assistant became a capable researcher.
And that’s the lesson for us business owners: AI isn’t dangerous. Vague leadership is.
Founders talk about delegation all the time, but most of us are bad at it.
We hand over chaos and expect brilliance—from employees, contractors, or now, robots.
The fix is the same across all three: clarity.
AI is the world’s most obedient, eager intern. It will do exactly what you ask—and throw in a few bonus surprises if you don’t give it boundaries.
If you tell it, “Go write something amazing,” it’ll proudly return nonsense and ask for a pat on the back.
But if you define the job, the scope, and the standards, it’ll exceed your expectations every time.
Lesson Learned. Going Forward I Will Be Clearer.
I almost embarrassed myself in front of my own daughter by sending a legally inaccutate letter to her landlord.
But the scare taught me a bigger lesson about leadership.
Business owners who avoid AI because it “hallucinates” are missing the point.
AI doens’t suck. How they treat it does.
Business owners need to treat it like an eager intern who can read your mind only if you teach it yours.
➡️ Learn how to train AI.
➡️ Learn how to talk to AI.
➡️ Learn how to correct it when it’s wrong.
➡️ Practice double-check its work (just like you would with any human)
And then it will be your best #1 🏆 employee that deserves every pat on the back you can give it.
🧠 Try This with AI
Copy + paste this before your next task to turn a hallucinating robot into your most reliable assistant:
Before completing my request, ask me:
1. What’s the goal?
2. Who’s it for?
3. What format do I need?
4. What tone or style should it use?
5. What facts must it check or verify?
Once I answer, restate my brief to confirm you understand—then begin.
💌 If This Hit Home…
Forward this to a friend who’s still afraid of “AI lies.”
Because maybe, like me, they just need better instructions.
Thanks for reading!
-Sonya
P.S. Before we could put up a big stink with the stinky landlord, my daughter found her keys in a coat pocket.
P.P.S. This is part of a series about AI for business owners. Next week I’m very excited to share my exact guide on how I train my AI for my business and then the next week I’ll share how to talk to AI (based on the latest research).