Safety and Speed Can Go Together

In October 1889, Judge magazine ran a cover called “An Unrestrained Demon.” It showed electricity as a giant spider. Its web was made of wires. A man hung dead in them while the crowd below ran for their lives.

The cover wasn’t made up. Two weeks earlier, a lineman named John Feeks grabbed what he thought was a safe telegraph wire above Manhattan. Blocks away, it had crossed a high-voltage line. His body hung in the wires for most of an hour while thousands of people watched.

The danger was real. But look at what happened next.

Nobody banned electricity. The city shut down the worst lines and forced companies to fix their wires and bury them underground. The fear was also bigger than the risk. Electricity caused fewer than 1% of accidental deaths in New York at the time. A few decades later, it was so safe and ordinary that people stopped thinking about it.

The pattern shows up again and again.

First, it’s a demon. Doctors warned that riding a bicycle would give women “bicycle face.” England made early cars crawl behind a man waving a red flag. People feared elevators would drop them to their deaths, and some did.

Then, someone built the brakes. In 1854, Elisha Otis stood on a raised elevator platform in front of a crowd and had the rope cut. His safety stopped the platform from falling. That one moment sold more elevators than any ad ever could. Otis didn’t ask people to trust the elevator. He showed them why they could.

Finally, it’s boring. Nobody thinks twice about riding to the 80th floor today. That is how you know a technology has won. People stop seeing it.

AI is in the demon stage right now.

Fear has followed almost every big technology. What feels different with AI is who is sounding the alarm. Some of the loudest warnings come from the companies building it.

That has happened before, too. In 1887, Edison’s company printed a pamphlet called “A Warning.” It was full of stories about deaths caused by alternating current, the system sold by his rival, Westinghouse. Edison’s allies then pushed cities to cap that current at 300 volts. The safety warning and the sales pitch were on the same page.

Now picture a frontier AI lab. Its biggest rival says its technology is moving so fast the world may need to slow down. What can it say back? It almost has to agree. If it disagrees, people hear one of two things. Either you don’t take the risks seriously, or your AI isn’t as good as theirs. Neither is easy to admit in public.

And there is a quieter effect. If you convince the world that AI is too powerful for ordinary rules, you also make the case that only a few large, well-funded companies should be allowed to build it. That raises the cost of competing for everyone else.

So one warning can do three jobs at once: real caution, strong marketing, and a path to rules that keep rivals out.

None of that makes the warning wrong. Some of the people raising it have nothing to sell. It just means you should ask who benefits before you decide how much weight to give it. Edison was right that high voltage wires could kill. He was wrong that the answer was to buy his system. The answer was better wires.

People keep asking if AI is moving too fast. None of us will settle that. The better question is one you can actually answer: Is your company moving faster than it understands?

Most are. Two in three people now use AI regularly. Fewer than half trust it. And 57% of workers say they hide how they use it.

Picture where that gap matters most. A nurse pastes a patient chart into a free chatbot. A banker asks an AI tool to sum up a client’s private file. Both are trying to do good work faster. Neither knows where that data went. Their bosses don’t either.

That is why hospitals, banks, and utilities move slowly with AI. They aren’t afraid of the technology. They fear what they can’t see. It is also why those places hold the biggest prize. The more sensitive the work, the more AI is worth once people trust it.

No lab and no law can make AI safe inside your company. That job is yours. It comes down to four questions:

1. Can we see it? Which tools, which people, which data.
2. Have we set limits? What can be shared, and what needs approval.
3. Who answers for it? The bigger the decision, the clearer the owner.
4. Is it paying off? Value created, measured against risk added.

A company that can answer all four can use AI in places its competitors are still afraid to go.

As investors, we read safety talk the way we read any letter from management. We listen to what it says, then look at what it protects. And when we study a company, one question tells us more than any other: would an employee describe its AI the same way the CEO does? When the answers match, management knows what it owns. When they don’t, risk is hiding somewhere. The few companies earning big profits from AI are about twice as likely to measure it. Good control and good returns tend to travel together.

As operators, we see the same thing from the inside. Every system we build goes into production, or we count it as a failure. Nothing reaches production without passing a security review. Safety isn’t the step that slows the work down. It’s the step that lets the work ship.

Every new technology looks like a demon at first. Some of the fear is real. Some of it is sold. Either way, fear was never the answer. Brakes were.

AI will make mistakes. So will the companies that use it. The winners won’t be the ones who avoid every problem. They’ll be the ones who see problems early, know who owns them, and fix them fast.

Otis didn’t ask people to trust the elevator. He showed them why they could. That is the job in front of every company using AI today.