Who Owns the Word "Safety"? What AI Regulation Teaches You About Reading the Hiring Game

I spent a Tuesday morning reading a press release I had no business caring about. A large company was applauding a government move that restricted a rival's most capable product. The language was clean: risk, responsibility, the public good. I almost scrolled past. Then I noticed, three paragraphs down and in nobody's headline, that the company doing the applauding had put billions into a third company that competed directly with the thing being restricted.

That is the whole story of AI regulation in one document, and it is also, strangely, the whole story of why you have sent 287 applications and heard back from four. Stay with me. The debate about whether AI rules are driven by genuine risk or by competitive advantage wearing a safety costume is the cleanest public example I know of a move that gets run on you, quietly, every week. The move is this: someone borrows the authority of a neutral-sounding rule to serve an interest that is anything but neutral. Once you can see it in the regulation fight, you cannot unsee it in your inbox.

This lesson is about ownership and borrowing. Who owns a rule, who only borrows its language, and how to tell the difference fast enough to stop taking the wrong rejections personally.

The difference between owning a rule and borrowing its voice

When you own something, you set the terms and you bear the cost of those terms. A landlord who owns a building can ban dogs and lives with the smaller tenant pool. The cost lands on the person who made the rule. That is ownership.

Borrowing is different. Borrowing is when you take a rule that exists to protect a shared good — safety, fairness, quality — and point it at whatever serves you, so that your private interest arrives dressed as a public one. The tell is always the same: the person invoking the rule does not bear its cost. Someone else does.

Watch how it works in the AI fight, because the financials are public and the shape is undisguised. A company invests heavily in one AI lab. That lab competes with another lab's model. When the rival model draws scrutiny and gets restricted in the name of risk, the investing company benefits twice over: its bet looks safer, and a competitor's product is off the board. The safety concern may even be sincere. That is the uncomfortable part. The point is not that the worry is fake. The point is that you cannot disentangle the worry from the windfall, and the people invoking the rule have arranged things so that they never have to.

This is what "follow the money" actually means. Not that everyone is lying. That incentives and language have fused so tightly you can no longer read the language straight. The word "safety" stopped being owned by anyone. It became a thing you borrow to win.

The same move, pointed at you

Now read your last three rejections again.

"We've decided to move forward with candidates whose experience more closely aligns with the role." "We're looking for someone who's a better culture fit." "Strong profile, but not quite what we need right now." Each of these borrows the voice of a neutral standard. Alignment. Fit. Need. They sound like ownership — like the company is bearing the cost of a careful, principled decision. They are almost never that.

I learned this the slow way. I applied to a mid-size fintech — not a name you'd recognize, a payments processor in the Midwest — and got the "not the right fit" line after a recruiter screen that had gone well enough that she'd talked about her commute. Two weeks later a friend of a friend got the same role. He had less experience than me on paper. What he had was a referral from someone three desks from the hiring manager. The "fit" rule wasn't owned by the company's actual needs. It was borrowed to ratify a decision that had already been made by proximity. Nobody who said "fit" paid a cost for saying it. I paid the cost, in the form of a polite email and a quiet afternoon.

Here is the thing that took me far too long to accept. The hiring system runs on borrowed authority the way the regulation fight does. "Bar raiser." "We hire for slope, not intercept." "We only want the top 1%." These are not lies exactly. They are private interests — speed, risk-reduction, the comfort of hiring people who look like the last person who worked out — wearing the costume of objective quality. And like the safety concern in the AI story, the worry is sometimes real. Sometimes the bar genuinely protects something. You still cannot read it straight, because the same words are used to mean "you're not good enough" and "we found someone's roommate" and "your resume didn't have the brand we were borrowing credibility from."

What the recruiter actually sees

Let me be specific about where the borrowing happens, because the abstraction is useless if you can't locate it.

A recruiter spends, by most honest accounts, somewhere between six and thirty seconds on a resume in the first pass. In that window they are not evaluating your skill. There is no time to evaluate skill. They are checking whether you carry borrowed credibility they can lean on when they pass you up the chain. A Stanford line. A Google line. A name their hiring manager will recognize. Those are not measures of how well you write code. They are tokens of trust that the recruiter borrows so that, if you turn out to be a bad hire, they can say "but the signals were there."

This is the engine of the FAANG pipeline, and it is why "learn to code, build projects, grind LeetCode, apply everywhere" never explained your silence. All five of those lessons assume the rule is owned — that there's a real standard, and meeting it gets you through. But the rule is borrowed. The recruiter is not measuring your ability against a fixed bar. They are looking for tokens they can borrow to protect themselves, and a self-taught developer from a state school is asking them to vouch with their own credibility instead. That is a much bigger ask than your resume admits, and nobody tells you that's the actual transaction.

I'm not saying this to make you angrier, though you have every right to the anger. I'm saying it because once you name the transaction correctly, you can stop trying to win a skill contest that isn't happening and start solving the trust problem that is.

Is AI regulation actually about safety or competition?

It is almost always both at once, and the people invoking the rule have an incentive to make sure you can never separate the two. That is the honest answer, and it is the answer you should carry into your job search.

In the regulation case, the safety argument and the competitive payoff travel in the same vehicle. You are shown the safety argument. The payoff rides in the trunk. The skill is not deciding "this is pure corruption" or "this is pure principle" — both are usually wrong. The skill is asking a colder question: who bears the cost of this rule, and who collects when it's enforced? If the person invoking it collects and someone else pays, you are looking at borrowed authority, no matter how sincere the speaker sounds.

Apply that exact question to a job rejection. Who bears the cost of "not a culture fit"? You do. Who collects? The hiring manager who got to make a low-effort decision and the referred candidate who skipped the line. The rule did not protect a shared good. It protected someone's convenience and someone's network, and it borrowed the language of standards to do it. Naming that is not bitterness. It is accurate reading. And accurate reading is the only thing that tells you where to spend your limited energy.

A small table for telling owned from borrowed

You can run this on a regulatory headline or on a rejection email. Same logic.

The vagueness row is the one I'd tattoo on my forearm. Borrowed authority hates specifics, because specifics expose the seam between the stated reason and the real one. A genuinely owned standard can tell you exactly what you missed — "we needed Kafka experience and you had none." Borrowed authority gives you "fit," "alignment," "not quite right," every time, because a specific reason would let you argue, and arguing would force the speaker to defend a position they don't actually hold.

What this changes about how you apply

I want to be careful here, because the easy move is to turn this into a grievance you nurse, and a grievance has never gotten anyone hired. The point of seeing the borrowing clearly is to stop spending yourself on the wrong fight.

Three things shifted for me once I started asking who owns the rule.

First, I stopped over-investing in the open application. If the front door runs on borrowed credibility I don't have, throwing more applications at it is throwing more resume tokens at people looking for tokens I can't produce. I cut my application volume by more than half and put the time into the side door — the engineers, not the recruiters. Not "networking" in the LinkedIn-spam sense. Asking a specific person a specific question about a specific problem their team had written about. That conversation makes me the borrowed credibility someone leans on. It flips the transaction.

Second, I started reading "fit" rejections as information about the company's incentives, not about my worth. When a company can't tell you a specific reason, it has told you something real: that its process runs on vibes and proximity, which means even a great interview was a coin flip you couldn't influence. That's not a verdict on you. It's a verdict on a door that was never fully open. You grieve it for an hour and you move to a company that can name what it wants.

Third — and this is the uncomfortable one — I got honest about where I was borrowing. I'd been writing cover letters full of borrowed language too. "Passionate about your mission." "Excited by the fast-paced environment." That's me invoking authority I didn't own, hoping the words would carry weight the substance didn't. The fix wasn't more polish. It was substance the recruiter couldn't get from anyone else: a paragraph showing I'd already used their product, found a specific rough edge, and had an idea about it. You cannot out-borrow the brand-name candidate. You can sometimes out-own them, by being the only applicant who did a thing nobody asked you to do.

The catch I have to name

There's a version of this lesson that curdles into "the whole system is rigged, so why try," and I won't sell you that, because it isn't true and it doesn't get you paid. Some standards are genuinely owned. Some companies really will tell you the specific skill you lacked and mean it. Some interviewers bear real cost when they pass on someone good and they take that seriously. The skill is not assuming everything is borrowed. It is being able to tell, quickly, which doors are which, and refusing to pour your one finite well of effort into the borrowed ones while telling yourself the problem is your work ethic.

The regulation story is unsettling for exactly this reason. You can't prove the safety concern was a costume. You also can't pretend the windfall wasn't sitting in the trunk. You hold both, and you act on the structure rather than the speech. That ambiguity is not a flaw in your reading. It is the actual shape of the situation, and people who refuse to sit in it get played by whoever is most comfortable borrowing.

Try this one thing this week

Take your last five rejections — or, if you can stomach it, the five job descriptions you're most intimidated by — and run a single column down a page. For each one, write the answer to one question: who bears the cost of this rule, and who collects when it's enforced?

That's it. Don't fix anything yet. Don't rewrite your resume. Just sort them. You'll find that two or three were owned rules — a real skill gap, a real requirement — and those are worth your effort, because effort actually moves them. The rest were borrowed, and the kindest thing you can do for yourself is stop trying to satisfy a standard that was never about you in the first place, and spend that hour finding the engineer who'll vouch instead.

The system will keep borrowing the language of fairness to do unfair things. You don't have to believe the language. You only have to read who's holding the trunk.