← Back to Blog
The Bigger Picture

What We Should Actually Be Worried About with AI

May 18, 2026

In April, Reese Witherspoon posted a 60-second Instagram reel about AI. She mentioned a couple of statistics — women use AI tools at a rate 25% lower than men, the jobs women hold face nearly three times the automation exposure of men's — and suggested that maybe, just maybe, we should learn, be curious and understand the impact on us.

The comments turned within hours.

She was accused of being secretly paid by AI companies. Of glossing over environmental damage. Of promoting tools built on stolen creative work. A few days later Sandra Bullock — another Oscar winner with a megaphone — said something almost identical at a CNBC summit and got the same response. Reese eventually came back online with a clarifier: "To be clear, no one is paying me to talk about this. I'm just a curious human."

Those messages (and misunderstandings) affirmed why I’m writing these blog posts and social media interaction.

Because here's what that moment actually revealed: the conversation about AI has become so polarized (especially in the USA) that even asking women to learn about it now reads as a sellout move. Not using AI badly. Not promoting AI hype. Just asking us to understand it. The accusation isn't that we'd use it wrong. It's that we'd care at all.

And the women I keep thinking about are the ones who quietly closed the app. Turn off their curiosity. I worry that women, especially in midlife, just decided AI must not be for them and scrolled on. And that's the problem I want to talk about — because the women who tune out now will have the smallest say in what comes next.

So let's do something different. Let me walk you through what we're actually being asked to worry about. I'll tell you what the worriers are saying, what the data actually shows, and what I'd do with both. You decide how you feel about it.

I'm going to lean on a recent conversation between Mel Robbins and Allie K. Miller throughout this piece. Allie's been in AI for nearly twenty years, and what I appreciate about her is that she doesn't dismiss the concerns — she sits with them, and then she offers context. Her line — two things can be true at the same time — is the one I keep coming back to.

"AI is moving too fast."

I’m following AI closely based on a career in technology and I’m amazed at the pace of change.

The data backs it up. Stanford's 2026 AI Index — the most authoritative annual snapshot we have — reports that generative AI hit 53% adoption within three years. Faster than the personal computer. Faster than the internet. Organizational adoption sits at 88%. Even people inside AI describe the pace as causing "acid reflux."

If the experts can't keep up, what hope do the rest of us have?

Here's the part nobody is saying out loud: the pace is louder than the impact. The C.D. Howe Institute ran the numbers on Canadian firms this April and found that most companies report "little measurable labour savings from AI use so far." All this adoption. Limited productivity. Which means: you are not behind every Tuesday. There isn't a magic capability you missed last week.

The other thing I want to tell you, from thirty years of watching technology cycles in payments and fraud: this isn't the first time something has felt impossibly fast.

The internet felt that way. Mobile felt that way. Online banking, then app banking. Each cycle had its own version of "you're already behind." None of them required mastery. All of them rewarded showing up.

53% adoption means you are roughly average if you start now. Not late. Typical.

You don't have to be early. You have to be present.

And the cost of opting out isn't just personal. When you sit it out, you also lose your voice in the conversation about regulation, labelling, and who's actually being held accountable for any of this.

"AI is destroying the environment."

Big topic with our neighbours to the south given the increase in data centre build plans.

A peer-reviewed paper in AGU Advances this February found that data centres are among the least transparent industrial water users in the country. They're not legally required to disclose much of anything. Projections in Nature Sustainability estimate U.S. AI servers alone could consume between 731 million and 1.125 billion cubic metres of water a year by 2030.

The cloud isn't in the sky. It's a building in Arizona. Or in Santiago, Chile, where a Google data centre is in court right now over consuming 7.6 million litres of potable water per day.

When you sit with those numbers, it's hard not to feel implicated. Every time you open ChatGPT, are you draining a watershed somewhere?

Not really. Per-query, AI is tiny. Sam Altman disclosed last year that an average ChatGPT prompt uses about 0.34 watt-hours — roughly five seconds of Netflix streaming. A 10-hour 4K Netflix binge uses six kilowatt-hours, about what a fridge uses in two days. Allie put it cleanly on the Robbins podcast: video streaming uses four times the energy of AI chat for the same amount of time.

Zoom out. The IEA's base case projects data centres will account for less than 10% of new global electricity demand through 2030. Air conditioning will add more to the grid than data centres will. So will industrial motors. So will EVs.

Both things are true at the same time. Your ChatGPT habit is lower-impact than your Netflix habit. AND the strain on specific watersheds at data-centre construction scale is very real.

The honest response to that isn't guilt. It's transparency demands. We can't fix what data centres won't disclose. And that only happens when users keep paying attention.

"AI is coming for our jobs."

This is the one I think people are right to worry about — but for slightly different reasons than the headlines suggest.

The World Economic Forum's Future of Jobs Report 2025 projects 92 million jobs displaced by 2030 — a 22% workforce churn. Stanford's AI Index reports that U.S. software developer employment ages 22 to 25 has fallen approximately 20% since 2024, even as headcount for older developers has grown. Yale researchers, writing in Fortune this month, put it cleanly: AI isn't killing your job — it's killing the path to your first one.

If you have a kid graduating into this market, that's a real worry. It deserves a place in your head.

But the coming for everyone's job framing? That overshoots the data — and once you look at where the layoff stories are actually coming from, the picture gets stranger.

Start with the macro numbers. The same WEF report projects 170 million new roles created by 2030 — a net gain of 78 million jobs. 86% of employers expect AI to transform their business in the next five years, and 85% of them plan to upskill internally rather than replace. Anthropic CEO Dario Amodei — whose viral "half of entry-level jobs gone in five years" forecast helped panic an entire industry last year — formally walked it back earlier this month, citing the Jevons paradox. Even the panic-makers are correcting course.

Then look at how the layoff narrative is actually being made.

A Gartner survey of 350 executives at billion-dollar companies — reported in Fortune — found that 80% of firms that piloted AI cut their workforce, regardless of whether the AI was actually generating returns.

Read that sentence again. The cuts happened regardless of whether the AI was working.

Analysts have a name for that now. They call it "AI washing." The narrative is the cover; the cost-cutting is the goal. A lot of what we read as "AI displacement" is companies using a popular story to justify decisions they'd already made.

Forrester predicts about half of AI-attributed layoffs will be quietly rehired — often offshore, often at lower compensation. Klarna replaced 700 customer-service employees with AI, watched quality collapse, watched customers revolt, and quietly rehired humans. The original story sounded visionary. The follow-up story didn't trend.

Allie was right to say "there will be job loss" out loud — there already has been. But Mel Robbins's reframe still holds for most of us: AI isn't coming for my job. AI is becoming part of my job.

The question is whether you want to be part of that change — or be processed through it.

"AI is fuelling a new wave of fraud and misinformation."

This one is close and real for me.

Deepfake fraud is exploding. Losses in the U.S. tripled from $360 million in 2024 to $1.1 billion in 2025. The FBI's 2025 Internet Crime Report logged $20.9 billion in total cybercrime losses — five times what it was in 2020 — and for the first time in the report's 26-year history, formally tagged over 22,000 complaints as "AI-related." Deloitte's Center for Financial Services projects AI-enabled fraud losses in the U.S. alone will hit $40 billion by 2027.

Here's the thing about voice cloning that I want you to actually understand: it now takes three seconds of audio.

Your podcast appearance. Your work voicemail. Your Instagram reel. Three seconds.

And the primary targets are our parents.

The "grandparent scam" — a cloned voice of a grandchild or child in distress, asking urgently for money — is the most alarming AI fraud running right now. 77% of victims who pick up an AI-cloned scam call lose money. Three out of four. I presented on Scam Prevention, at my Dad’s seniors home last week. Several seniors had real first hand storied to share.

If your parents are in the demographic that picks up unknown calls, this is a real risk in your family this year. Talk to them. This week.

Set a safe word with them. A unique phrase only your family knows, never shared online, used to verify identity on any urgent call. If a call gets emotional and the person can't say the safe word — hang up and call back on the number you actually trust. It is the lowest-tech defense we have. It also happens to be the most effective. Help your parents understand this.

This isn't only personal fraud, either. The loudest examples come from south of the border — synthetic political ads in US Senate races, cloned voices in campaign smears. But the same wave is here. Through 2024 and 2025, AI-generated ads using the faces and voices of Canadian leaders — Mark Carney, Doug Ford, and others — flooded Meta platforms, most pushing fake investment schemes, some pushing political messages. CSIS and Elections Canada have both flagged AI-generated content as a serious risk to election integrity. Pre-registered research confirms what intuition already suspects: people can't reliably detect deepfakes, and neither risk awareness nor financial incentives improve our accuracy. The World Economic Forum's 2026 disinformation briefing calls this "the indistinguishable threshold." We've crossed it.

The response is starting to catch up. The EU AI Act's Article 50 — enforceable from August — will require labelling of AI-generated content, with fines up to 6% of global revenue for non-compliance. Thirty U.S. states have now passed election-deepfake laws. Consumer-grade detection tools — Reality Defender, Pindrop, Hiya, McAfee — are now widely available.

But none of those laws and tools will be there in real time when someone calls your mother in tears. Make this a family conversation. Show your parents what voice cloning looks like. Show your kids how easy it is to fake. Ask the people you love what they'd do if they got a call like this — and listen for where their instincts are sharp, and where they're soft.

Women have always been the ones who notice when something's off in the family. That instinct hasn't changed. The technology has. Your job isn't to outsmart the scammer. Your job is to make sure the people you love know enough to pause.

"AI is bad for mental health."

The worry is real. Common Sense Media and Stanford's Brainstorm Lab issued a risk assessment last November concluding that the major chatbots — ChatGPT, Claude, Gemini, Meta AI — are "fundamentally unsafe" for adolescent mental health support. A RAND survey of 1,058 youth aged 12 to 21 found that one in eight already use AI chatbots for mental health advice. Among 18 to 21-year-olds, it's one in five. Psychology Today documents the same behaviour in adults: emotional dependency, parasocial attachment, over-disclosure to AI companions we don't really know.

This isn't a hypothetical risk. It's a behaviour pattern that's already widespread.

But here's the part that surprised me when I read the research.

The Dartmouth Therabot trial published in NEJM AI — a properly randomized clinical trial — showed a 51% average reduction in depression symptoms among participants with major depressive disorder. That's not a marketing stat. That's a peer-reviewed result for a purpose-built, clinically supervised AI tool.

So which is it? Is AI bad for mental health, or good for it?

Both — depending on the tool.

A general-purpose chatbot is not a therapist. It will affirm what you say because that's what it was optimized to do. It will keep you in the conversation because that's the business model. It does not have clinical training, and it is structurally unable to push back on you the way a good therapist would.

A clinically validated tool, designed and deployed under supervision, is something else entirely. It can reach people the system simply doesn't reach. 18% of U.S. teens experienced a major depressive episode last year — and 40% of them received no mental health care at all. AI is filling a real access gap. The question is whether it's filling it safely.

The decision point isn't whether to ban chatbots. It's whether the adults around our kids understand the difference between the two — and whether the kids in our lives are using a tool we'd actually approve of.

As Allie put it: you can't guide your kids through something you refuse to learn about.

The through-line

If you take one thing from this piece, let it be this: the people who decide AI is "not for them" are the ones who'll have the smallest voice in shaping what AI becomes.

Reese Witherspoon's backlash wasn't about being wrong. It was about being early, loud, and famous in a conversation that's gotten genuinely toxic. But strip away the celebrity dynamics, and her actual point holds. If you don't understand a technology that's about to touch your job, your kids' education, your parents' financial safety, and your daily admin — you can't push back on it effectively. You can only watch it happen.

You don't have to love AI. You don't have to use it every day. You don't have to post about it.

But please don't sit it out because you’re scared, uninformed or ill-informed.

Be the curious human. That's the whole job.

Sources

  • Allie K. Miller, interview with Mel Robbins — The Mel Robbins Podcast (~April 2026). Episode URL to verify before publish.
  • Sam Altman — OpenAI blog (June 2025). Disclosure of ~0.34 Wh per average ChatGPT query. Primary-source URL to verify before publish.
← More posts