My daughter Raeleigh is convinced artificial intelligence is eventually going to kill us all. She is also an alarmist, and even she knows this about herself. Rae has never met a possibility she could not immediately carry to its most dramatic conclusion. If the weather app says thunderstorms, she is halfway to imagining the tornado. If somebody online says a food additive might be questionable, she is probably wondering whether our organs are already shutting down.

She is extreme. I love her for it. She also makes the rest of us a little crazy.

So when I saw a headline this week about an AI researcher quitting his job and warning that advanced AI could eventually become dangerous enough to threaten humanity, I stopped and read it. Mostly because I knew Rae would too, and I wanted to know what was actually being said before she showed up ready to explain that humanity had six months left.

The concern was about moving toward increasingly powerful systems faster than researchers can figure out how to reliably control them. The part that stuck with me was the race. If one company slows down, another may keep going. If one country decides to be cautious, another may see that as an opportunity to pull ahead.

That felt familiar. People do this all the time. We know slowing down would probably be wise, but we worry about what we will lose if everyone else keeps moving. We know there is risk, but we tell ourselves we can deal with it later. We wait for more proof, more certainty, one more sign. Sometimes the cost of waiting gets much higher than we expected.

When Raeleigh was born, I was told at the hospital that I had a sexually transmitted infection. I knew I had not slept with anyone except my husband. I also knew I had tested clear for sexually transmitted infections at the beginning of my pregnancy. There were not many conclusions available after that.

I confronted Tyree. I told him I knew and that this was his one chance. For our daughter, I was willing to stay and try to make our marriage work. I told him I would not tolerate it happening again.

I wanted another explanation to exist. I had just had this beautiful baby. I was happy. I wanted her to have the family I thought I was giving her. Maybe pregnancy hormones had something to do with it. Maybe hope did. Maybe I just was not ready to blow up my life over something I already knew in my gut was true. So I stayed for three more years.

During those years, life kept moving. We were still a family. There were still ordinary days and kisses goodbye and all the routines that make it easier to convince yourself maybe the thing you fear is not really happening after all.

Then one day, I had just left Tyree and was driving with Rae in the back seat. He had kissed me goodbye before we got into the car. Rae was strapped into her booster behind me when she asked, “Mommy, why does Daddy kiss Tia like he kisses you?”

I looked at her in the rearview mirror.

Eventually I learned there were two women. That was the end.

I learned a lot from that marriage, but one of the hardest lessons was that I had known enough much earlier than I wanted to admit. I kept waiting for certainty because acting on what I already knew would have meant making a painful decision while I was still hoping I could avoid it.

That is part of why I think about safety the way I do now.

I make small decisions about risk with my kids all the time. Most of them do not look dramatic from the outside. We call them accommodations.

Royce has a tendency to bolt when he gets upset. I also have mobility limitations, which means I cannot count on being able to chase down a determined toddler if he takes off toward something dangerous. So when I am alone with him, I sometimes use a backpack leash.

People have opinions about those. They had opinions when Rae was little too. “It’s a child, not an animal.”

Thank you, stranger. I had somehow missed that.

When Rae and my nephew Sylas were little, I often had both of them with me. My knees were already bad, and those two were perfectly capable of deciding to go in opposite directions at the exact same moment. The leashes meant they could walk, explore, and stay on their own feet without me having to keep them strapped into a stroller because I was afraid I could not physically keep up.

Now the same kind of backup helps me with Royce. I still hold his hand. I still walk beside him. I still teach him to stay close. The leash is there in case those things fail. It accommodates his elopement risk and my mobility at the same time, and most importantly, it means we still get to go.

That is what I think people sometimes misunderstand about guardrails. They can create freedom. A safety feature can be the reason participation is possible in the first place.

The same principle is why I buckle my child into a car seat every time we drive. I do not need a wreck to prove the seat was worth using. The possibility alone is enough when the consequence is that serious.

That is where I keep landing with AI too.

No, Rae, I do not think the robots are coming to kill us tomorrow.

I do think there are real concerns that deserve serious attention. Powerful systems should be tested carefully, and countries need shared rules so caution does not become a disadvantage in a global race.

The part that worries me most is how ordinary the pressure to keep going can be. Companies want to stay ahead. Countries want to stay ahead. Nobody wants to slow down while imagining everyone else sprinting past them.

That is where the comparison to the atomic bomb makes sense to me. The technology is different. The race itself is familiar.

And races like that are exactly why safety cannot depend on everyone independently deciding to be responsible. Some problems require shared guardrails before anyone has crossed the point where they matter most.

There are plenty of mistakes we can afford to make. Build the website wrong. Burn dinner. Take the job and realize you hate it. Paint the room an ugly color. Make something badly and rebuild it better. Trial and error works when there is another trial.

Some consequences do not give you that. A toddler running into traffic is not something I am willing to use as a learning experience. A fatal crash is not a lesson anyone needs in order to appreciate the value of a car seat. If artificial intelligence ever reaches the point where it can cause truly catastrophic harm, I do not want humanity to finally decide at that point that maybe guardrails would have been useful.

I have learned plenty of things the hard way. Sometimes because I had no choice. Sometimes because I did not know better. Sometimes because I knew enough to see the warning and still hoped reality would give me a different answer.

These days, I have much more respect for the boring little things that keep disaster from becoming a lesson. The car seat. The backpack leash. The extra test. The rule everyone agrees to follow. The warning someone chooses to take seriously before they have absolute proof.

Safety means looking honestly at what could go wrong, deciding what the cost of being wrong would be, and putting something in place before that cost comes due.

Some lessons can be learned by screwing up and trying again. Some are too expensive for that.

And if the only thing left to say afterward is, at least now we know, sometimes we waited far too long.