You Own Them

Hey everyone,

What if the warning is less accountability than a credential?

I am getting really fucking tired of watching some of the most powerful people in AI perform this bizarre ritual where they confess to building something potentially dangerous and somehow expect the confession itself to make them more credible. You know the routine by now. We built something incredibly powerful. Maybe too powerful. We moved incredibly fast. Maybe too fast. There are risks we don't fully understand. Some of them could be catastrophic. We tried to warn people. And now, having bravely admitted all of this, we'd like governments, regulators, and the general public to listen very carefully while we explain what everyone else needs to do about the problem we created.

What the fuck?

Imagine this in almost any other industry. Imagine the CEO of a pharmaceutical company walking onto a stage and saying, essentially, "We developed a drug that we now believe could seriously injure or even kill people. We weren't entirely sure what it would do, but the market opportunity was enormous, so we kept going. We distributed it widely. We raised billions of dollars. We competed aggressively with anyone trying to beat us to it. And now we've discovered that this might be dangerous." You would not call that person brave. You would call an attorney.

And yet somewhere along the way, parts of the AI industry developed this remarkable moral loophole where publicly acknowledging the potential consequences of your own decisions is treated as evidence that you should be entrusted with even more authority over what happens next. I don't buy it. If you genuinely believe the technology your company is building presents an unacceptable danger to the public, then act like it. Slow it down. Don't ship it. Don't sell it. Accept the financial consequences. Resign if you believe the company is behaving irresponsibly and you cannot stop it. Testify about what happened. Bring receipts. Take responsibility. But don't spend years racing toward the frontier because the rewards might be enormous and then, once the consequences become harder to ignore, suddenly discover the comforting embrace of government intervention.

Leadership does not work that way. You don't get unlimited upside while the experiment is going well and socialized downside when it isn't. And that's the part of this entire conversation that makes me furious, because ordinary people did not make these decisions. A schoolteacher didn't decide how quickly frontier models should be deployed. A truck driver didn't set the safety budget. A graphic designer didn't decide whether competitive pressure justified another release. A family trying to pay their mortgage didn't choose the training runs, the deployment schedules, the safeguards, the corporate structures, or the incentives. Executives did. Boards did. Investors did. Researchers did. Companies did.

That doesn't mean every one of those people behaved recklessly, and it doesn't mean every warning is cynical. There are researchers and employees raising legitimate concerns, sometimes at considerable professional cost, and those concerns deserve to be heard. But there is a difference between warning the public and laundering responsibility through the warning. "I helped build this and I think there is a serious problem" should be the beginning of accountability, not the end of it. And it sure as hell should not become a credential.

I keep hearing versions of the argument that AI is simply too important, too complicated, too globally competitive, or too dangerous for ordinary standards of corporate responsibility to apply. I think the opposite is true. The more powerful the technology, the greater the responsibility of the people building it. That's the fucking job. If you're going to stand onstage and tell the world that you're building one of the most consequential technologies in human history, then congratulations: you've also volunteered for one of the largest burdens of responsibility in corporate history.

You don't get the first sentence without the second.

You don't get to tell investors you're building the future and tell regulators you couldn't possibly have anticipated what happened. You don't get to describe yourself as the steward of civilization when things are going well and a helpless passenger when they aren't. And you definitely don't get to fuck around at unprecedented scale and then hand the bill to everyone else when the auditors start asking questions.

Maybe some of these warnings are exactly right. Maybe some risks are worse than the public understands. Maybe governments genuinely need to act. Maybe we need much stronger safety standards, transparency requirements, liability regimes, independent testing, or entirely new regulatory structures. Fine. Then let's have that conversation. But start with the oldest governance mechanism humanity has ever invented: responsibility. Who made the decision? Who benefited from it? Who knew what, and when? What safeguards existed? What failed? Who signed off? And who bears the consequences?

Because there is something deeply corrosive about a culture in which enormous power comes with enormous compensation, enormous prestige, enormous influence—and apparently no corresponding obligation to own the outcome. That's not leadership. That's a clown car with a cap table.

And maybe that's the thing bothering me most. AI itself still fascinates me. I build with it every day. I think it can expand what human beings are capable of creating, discovering, and understanding, and I think there are extraordinary things ahead of us. Which is precisely why this matters. I don't want the future of this technology decided by people who believe accountability is something you outsource after the Series C. I want builders who understand that stewardship isn't a press release you issue when things get uncomfortable. It's what you practice while you're building.

Build boldly. Absolutely. Take risks. Explore strange territory. Push against what everyone else thinks is possible. But if you're going to build boldly, you also have to observe honestly. And when what you observe tells you that something you've built is dangerous, broken, or causing harm, you don't get to turn around and pretend the consequences belong to somebody else.

You own them.

That's the deal. Or at least it fucking should be.