"Move Fast and Break Things" Breaks Things for Women: Why Equality Needs Regulation With Teeth.
- Jun 27
- 8 min read
Updated: 4 days ago
A reader at a recent fireside chat asked me what I thought about the European Union's AI Act, the law that puts real penalties on AI systems that discriminate. I said I was here for it. All of it. And it reminded me of a book I read, one that I could not put down, and realized the law and the book were telling me the exact same story.
The book is Careless People by Sarah Wynn-Williams. If you have not read it yet, move it to the top of your list. She spent years inside Facebook, now Meta, as a director of global public policy, in the rooms where the biggest decisions about the biggest platform on earth got made. Her account is a first-hand look at what happens when a company chases growth and power and government access, and treats the human cost as a rounding error.
I am not going to spoil the book for you, but I am going to tell you why, as the author of The AI Gap and someone who spent 15 years inside corporate systems as a Chief People Officer, I think every ambitious woman should read it through one specific lens. Because the carelessness she documents is not history.
It is the operating model of the companies building the AI that is screening your resume, ranking your promotion, and deciding who gets seen right now.
WHAT "CARELESS" ACTUALLY MEANS
Here is the line from the book that everything else hangs on. Wynn-Williams borrows the idea from The Great Gatsby, the part about careless people who smash up things and creatures and then retreat back into their money and let other people clean up the mess.
That is the whole thesis. Not evil people. CARELESS ones.
There is a meaningful difference, and it matters for how we respond. Evil is rare and obvious and easy to point at.
Carelessness is common. It is the executive who is not trying to harm anyone, who is simply not slowing down long enough to ask who gets hurt when the thing scales. It is growth treated as the only goal that counts. It is "we will fix it later" said by people who never have to live inside the later.
And here is why I am telling you this in a piece about AI.
The same companies. The same playbook. The same posture of "trust us, we will handle the consequences." Except now the product is not a social feed. It is the decision-making infrastructure of the entire economy.
WHY I STOPPED TRUSTING "WE'LL SOLVE IT UPSTREAM"
When I talk about The AI Gap, the most comfortable thing I hear in response goes something like this. "The bias will get solved upstream. The smart engineers will handle it. The rest of us just inherit a cleaner system eventually."
I used to want to believe that. I do not anymore. Careless People is a 300-page explanation of why.
Because the question was never whether the people building these systems are smart. They are brilliant. The question is whether anyone is structurally required to care about the harm. And what Wynn-Williams documents, again and again, is what happens to the people inside who raise their hand and say "this is going to hurt someone." They get managed out. They get sidelined. The growth target wins and the warning loses.
You cannot reset a baseline you are standing inside of. And you cannot count on a company to police a harm that costs it money to fix, when its entire culture rewards the people who look away.
THE DEFENSE DEPARTMENT PILOT THAT SHOULD HAVE YOUR ATTENTION
Let me bring this fully into 2026, because this is the part that I worry about.
Over the past year, the AI companies h
ave raced into government contracts. OpenAI signed a pilot with the U.S. Defense Department reported at a $200 million ceiling, and a separate deal through the General Services Administration that makes ChatGPT Enterprise available to nearly every federal agency for one dollar a year. One dollar.
They are not in it for the dollar. They are in it for the position. Get the technology embedded inside the most powerful institutions on earth, and you become infrastructure that is very, very hard to remove.
Now here is where everyone should be paying attention:
According to reporting from TechCrunch and the companies' own published statements, when these defense deals were being negotiated, the labs were not all willing to draw the same lines. OpenAI moved forward with a contract allowing its models to be used in the military's classified network. Anthropic, the maker of Claude, refused to allow its technology to be used for mass surveillance of Americans or to power fully autonomous weapons. For holding that line, Anthropic was reportedly named a supply-chain risk by the Department of Defense and has since sued the Pentagon.
I am not here to crown a hero. Every one of these companies is a business with its own interests, and you should trust none of them blindly. But sit with the shape of that story for a second.
One company drew a red line on surveillance and autonomous weapons. And drawing the line is what got it labeled a risk.
That is Careless People in real time. That is the exact dynamic Wynn-Williams documents, just with higher stakes. The person, or in this case the company, that says "wait, who does this hurt" becomes the problem to be managed, while the one that says "yes, ship it" gets the deal, the access, and the position inside the machine.
When the screening tools and the surveillance tools and the weapons systems are all being built by the same handful of companies, by the same handful of people, under the same pressure to win the contract and worry about the consequences later, you do not get to assume someone responsible is minding the harm. You have to require it. In writing. With teeth.
WHAT THE EU AI ACT HIRING RULES ACTUALLY DO
So let me actually answer the question I got asked in that fireside chat.
The EU AI Act is the first serious attempt by any government to require care instead of hoping for it. Here is the plain-English version.
It sorts AI by risk. The more a system can hurt you, the stricter the rules. Some uses are banned outright, like AI that scores people by their behavior or reads workers' emotions on the job. Some are low risk, like spam filters, and barely regulated. And then there is the category that should matter to every woman reading this.
HIGH RISK. And do you know what sits squarely in the high-risk bucket? AI used for recruitment, resume screening, candidate ranking, performance evaluation, promotion, and termination. The exact systems creating the AI Gap I wrote a whole book about.
Starting August 2, 2026, companies using those systems for anyone in the EU have to do the things careless companies never do on their own. Test for bias. Document how the system works. Keep a human meaningfully in the loop. Tell you when AI is being used on you. And give you the right to an explanation when an algorithm makes a decision about your career. The penalties are built to be felt, up to 35 million euros or 7% of global revenue for the worst violations. That isn't a slap on the wrist, but something that has meaningful impact to their balance, should they violate the act.
And, if there is one thing I've learned as a Chief HR Officer, it's quite funny how fast things move when they hit the balance sheet.
THE TELL: EVEN THE FLOOR IS GETTING LOBBIED
Here is the part that proves my whole point about accountability, and it is happening as in real time.
That August 2026 deadline is under pressure. A proposal called the Digital Omnibus surfaced in late 2025 trying to push the high-risk hiring rules all the way back to December 2027, and as of this spring the negotiations had not cleanly resolved. So the floor, the bare minimum, the "you must test your hiring tool for bias" rule, is itself being lobbied against before it ever takes effect.
Don't let that be a side note, here. That IS the story. Accountability is always the uncomfortable part. Nobody throws a party for the audit. Nobody lobbies to be regulated. The careless people do not announce themselves (And TBH - they probably don't view themselves as careless. They probably view themselves as pioneers, driven, visionaries, etc. They just quietly push the deadline, water down the language, and hope that you are not paying attention.
So, please, pay attention.
WHY REGULATION IS THE FLOOR, NOT THE CEILING
Now, do I think a European law fixes this for all of us? No. Regulation is the bare minimum, not the ceiling. It buys us a little time. It raises the cost of carelessness. But it still does not put a single additional woman in the room where these systems are designed.
Because here is what both the book and the law point to. The problem was never only the technology. It was who got to be careless with it, and who had no seat at the table to say - wait a minute dude - hold up. When the senior technical teams at the major AI labs were essentially all men in the first drafts of this era, and have only crept up to roughly 10% women at the top since, the carelessness is not random. It runs in one direction. It always has.
A law can punish the harm after the fact. Only representation prevents it on the front end. We need both.
We need the policy with teeth AND we need women at every rung of that ladder, building, challenging, refusing to look away, pulling each other up into the rooms where the red lines actually get drawn.
WHAT I WANT YOU TO DO WITH THIS
Read Careless People. Treat is as a field guide. Learn what carelessness looks like from the inside so you can recognize it when it is dressed up as innovation. (Trust me, it's a fascinating read, and I'll probably come back to it again after writing this post!)
Then refuse to be managed by it. Run the prompt that shows you your own bias in these systems and challenge the output out loud. Ask your employer what AI is being used on you in hiring and reviews, and whether anyone has tested it. Get yourself fluent enough that you are in the room, not just subject to it.
The careless people are counting on us to stay polite, stay grateful, and stay quiet while they push the deadline one more year.
I am not built that way.
And neither, I suspect, are you.
Frequently Asked Questions
**What is the EU AI Act and who does it apply to?**
The EU AI Act is the world's first comprehensive law regulating artificial intelligence. It sorts AI systems by how much harm they can cause and applies the strictest rules to high-risk uses like hiring, promotion, and performance evaluation. It applies to any company whose AI affects people in the EU, even if that company is based in the United States.
**Does the EU AI Act cover AI used in hiring and promotions?**
Yes. AI systems used for recruitment, resume screening, candidate ranking, performance evaluation, promotion, and termination are explicitly classified as high-risk. Starting August 2, 2026, companies using them must test for bias, keep a human in the loop, disclose when AI is being used, and give individuals the right to an explanation, with penalties reaching up to 35 million euros or 7% of global revenue for the most serious violations.
The AI Gap is my second book, about how AI systems trained on biased historical data are quietly widening the gaps women have spent decades fighting to close, in pay, leadership, visibility, and opportunity. It argues that closing the gap depends on women getting fluent in AI and into the rooms where these systems are built, not waiting for the engineers to fix it for us.
Comments