Why I Care That You Don’t Know How to Use AI

I asked the room, again, if anyone had heard of AI hallucinations. The silence that followed was the kind of silence I have come to recognise as a Territory specialty: composed, ancestrally patient, entirely impervious to follow-up. It is not hostile. It is not even uninterested. It is the silence of a population that has survived cyclones, federal funding cuts, and three rebrands of the local council, and has decided that whatever I am selling can wait. Normally I respond to this silence with professional good cheer. This time, something inside me curdled.

The Silence I Have Come to Recognise, and the Curdling That Followed It

I had opened the workshop, as I always do, with the gentlest possible diagnostic question: who in this room has used ChatGPT, Claude, or Gemini? Two hands. Both belonging to the same woman, who had also brought biscuits. I took this as a soft warning and moved on. Then I asked, slightly more pointed: has anyone here heard the term “AI hallucinations”? And there it was. The silence. The familiar, weather-system silence that descends in Pandanus Reach training rooms whenever the topic threatens to require an opinion.

I have, over the past year of delivering this material, developed a strong professional tolerance for this silence. I let it sit. I smile. I redirect. I produce, from a back pocket, a small definition I have rehearsed for exactly this purpose. It usually works, in the sense that the workshop continues and nobody cries.

But this time, I felt something new behind the smile. Not annoyance, which would be unprofessional. Not condescension, which would be unkind. Something more interesting: a tight, private knot of self-interested resentment, dressed for public viewing in a cardigan it had borrowed from professional concern. It had been there for a while. It had only, until that morning, been wearing a different cardigan.

A Confession Concerning the Birthday-Text Industrial Complex

To explain what this feeling is, I have to take you on a short detour through my contacts list.

The people who do not know what an AI hallucination is are, with a precision that ought to alarm somebody, the same people who possess large quantities of my personally identifiable information. My phone number lives on their phone. So does my email. So does, in many cases, a contact photo of me from 2017 that I would not have chosen. Their iPhones, the older of which are running an operating system released two prime ministers ago, have logged my home address as a “frequent location.” I have, by virtue of being a human being with relatives, distributed myself across a small constellation of devices over which I have no jurisdiction and concerning which I have, increasingly, no peace.

These are also, and this is the comedy of it, the same people who text me at 6:47am on my birthday. Not because they remembered. Because their phone reminded them, using a contact card that contains my full legal name, my mobile, my email, and possibly the date I last visited their house, which iOS has been quietly logging in the background since approximately the Rudd government. They love me. They also have me, in data form, sitting in a device whose lock-screen passcode is the year of their wedding, which is on their Facebook.

I have, in the course of my professional career, developed a quiet and not entirely virtuous concern about other people’s digital habits. It presents publicly as I just want you to be safe online, love. Internally, it is closer to: you are the soft underbelly of my own digital security, and I have, on at least three separate occasions, fantasised about throwing your phone into the harbour.

In the privacy of my own thinking, I have more than once imagined a small, dignified intervention in which I gently but firmly remove a phone from a loved one’s handbag and replace it with a brick. The brick would be safer. The brick has never clicked on a parcel notification.

I do not think I am alone in this. I think most people working in or adjacent to information security harbour a version of the same dark, self-interested concern, which they frame, with great social discipline, as helpful advice. It is helpful advice. It is also, underneath, a recognition that the person in question is a liability not only to themselves but to everyone whose number lives in their phone.

An Incomplete Lexicon of Things We Say When We Mean Something Worse

In the interests of public service, and in the spirit of bringing a problem into the light where it might be more easily contained, I offer here an incomplete lexicon of phrases used by cybersecurity-adjacent professionals when addressing beloved family members whose digital practices have begun to give them a quiet, persistent stomach ulcer.

The phrases are gentle because the relationships are non-negotiable.

The Drunk Driver Offers You a Lift

Here is the analogy I have been carrying around for a while and have, until this morning, kept to myself for reasons of social hygiene. (See above.)

The cybersecurity-illiterate friend is, functionally, the drunk driver who has just offered you a lift home. They are not malicious. They are warm. They are insisting. They have already picked up your handbag. They are pointing at the carpark with the keys held aloft in a way that suggests the offer has, some minutes ago, ceased to be a question. And so you sit in the passenger seat, white-knuckled, watching them merge onto a freeway of phishing emails, unencrypted Wi-Fi, and SMS notifications from companies you have never heard of, asking very gently, with the cadence of someone who would like not to upset them: are you sure you’re right to drive?

And they say: I’ve been doing this for years, love. She’ll be right.

And technically, statistically, they have been doing this for years, and it has, in fact, been right. Until the Tuesday afternoon when a credential they reused on a forum in 2014 surfaces on a paste site, and the chain of unpleasantness that follows includes, among other things, you, because your number is in their contacts, and your maiden name is in their messages, and the recovery email for their bank login is, on closer inspection, an address you co-administered with them in 2011 and forgot existed.

Now Replace the Car With the Internet, and the Drunk Driver With Approximately Everyone

The AI-illiteracy version is the same drunk driver. There are simply more of them, and they are not slowing down for the corners.

They cannot tell that the image of the politician was generated last Thursday by someone in a different hemisphere with a graphics card and a grudge. They cannot tell that the news article that confirmed everything they already believed was hallucinated by a chatbot rewarded, during training, for sounding confident. They cannot tell that the email from the bank is not from the bank, because the bank’s actual emails also sound like they were written by a chatbot rewarded for sounding confident, and which therefore reads as more legitimate, not less. The bar for plausibility has been raised by the very technology that makes the implausible easier to produce. Discernment, as a category, is becoming a luxury good.

They send the article to me with a row of exclamation marks. They reply to the bank email with their account number, because the email said it was urgent. And the next day they get on with their lives, which intersect, in a great many places, with mine.

This is the bit that makes me want to leap into their handbag.

The Serious Bit, Which Is, This Time, Genuinely Serious

I am going to drop the comedy for one paragraph. It will return.

Digital literacy is not a personal preference. It is not yoga. It is not whether you enjoy oat milk in your flat white or have strong opinions about your local council’s parking strategy. It is closer to driving, or vaccination, or stopping at a red light. The thing you do, or fail to do, on the internet does not stay on your own little patch of internet, because the internet is, and has always been, a network of interconnected computers. The clue is in the name. We are not each living in our own private metaverse. We are sharing one set of pipes, one set of platforms, one fraying epistemic commons; and the person sitting next to you on the Pandanus Reach bus, reading a fabricated article about a politician on a website that did not exist six months ago, is going to vote in the same election as you, drink from the same water supply as you, and forward the article, with conviction, to a group chat that contains your mother.

That is the externality. It is unfashionable to name. It is also true.

What I Actually Want to Say in Workshops, and What I Say Instead

So when you sit in my workshop and tell me, with the soft pride of a generational hold-out, that you “don’t really do AI,” I smile politely. I nod. I say something professionally encouraging about meeting people where they are. I scroll to the training materials section titled What Even Is a Large Language Model, Slowly, With Pictures.

What I am doing internally is calculating the blast radius.

I am not calculating it punitively. I am calculating it because I have to. Your inability to identify an AI-generated photograph affects what photographs will, in due course, be shared with me. Your inability to spot a fake website affects which websites will be linked to me, by you, with a kind instruction to read the article. Your inability to understand what AI can and cannot reliably do affects whose advice you will follow, whose products you will recommend, and whose deep-faked voice on the phone, claiming to be a grandchild in trouble, you will believe long enough to drive to the chemist for a gift card. (This last scam is real. It is running, today, in Australia. The relevant consumer-protection agencies have had words about it, and the words have not, as yet, fully helped.)

This is not a private matter. It has never been a private matter. I am genuinely sorry to be the one to tell you.

I do still love you. I would still come to your birthday. I would just, very gently, like to take your phone.