Inside Silicon Valley: The People Building AI Are Worried Too

“I cried at my desk a lot those days. I remember going from over 200 people in the company to 30. It was brutal.”

So often, when we talk about AI, we hear from the people standing outside of it.

The family living beside a new data centre. The graduate struggling to find work. The CEO who has frozen hiring. The economist predicting what automation will do to the labour market. The environmentalist warning about energy consumption. The artist and writer wondering what happens when a machine can produce their work in seconds.

It is undeniable that AI has irreversibly affected every single one of our lives. And it is important we speak up.

But what about the people inside?

Behind the polished offices and carefully managed public statements of OpenAI, Anthropic, Google DeepMind, and the rest of the rapidly expanding AI industry are the people actually building it. People tasked with developing increasingly powerful systems while simultaneously trying to work out what those systems should and should not be allowed to do.

And what about the people who walked away?

People who left jobs that could pay six, sometimes seven figures because their concerns about what they were helping to build became too difficult to ignore. People who became frightened by the direction of the technology, frustrated by the industry’s approach to safety or simply decided that they no longer wanted to be part of it.

I wanted to hear from them.

But I also wanted to hear from the people who stayed.

The employees trying to push their companies towards greater safety. The people raising concerns from inside the organisations developing the technology. The people who believe the industry can still be changed, and who are trying to change it from within.

Because if we want to understand where AI is taking us, perhaps we should start by asking the people who are already inside.

All anonymous sources have been independently verified. Their identities have been withheld to protect their employment and personal privacy. Claims have been independently assessed and relevant allegations put to those concerned for response.

The AI Mandate: Build It Faster

“It was not an experiment. It was a mandate: build what I want faster, or else.”

Across the interviews I conducted, one instruction appeared again and again: build it faster.

One person, who spent almost a decade working in Silicon Valley, described the immense and consistent pressure placed on software engineers to produce new products quickly, even when they believed those products were not ready. The pressure, she told me, came from above.

Teams were expected to keep moving, with deadlines and productivity taking priority over whether what they were building was actually ready to be put in front of users, or even safe and ethical.

“I did not want to send out slop quickly to make the bosses happy about meeting an imaginary deadline.”

For her, that meant watching considerations she believed should be fundamental to building software pushed aside in the rush to get products out. She describes basic safeguards being overlooked, including features that would allow users to report problems or raise concerns about what they were experiencing.

“I cared about the users of my creations, and I wanted to take care of them.”

That, she says, was increasingly at odds with the environment she was working in.

“My moral and ethical compass says no.”

“But we were made to keep moving forward at an unreasonable pace for the sake of the shareholders.”

The issue, then, is not simply that some workers dislike working quickly. It is what happens when speed becomes the priority even when the people building the product believe something is not ready.

And for some, that pressure was beginning to change what it meant to work in Silicon Valley at all.

The Culture Shift: Alienation in Silicon Valley

“I went from respecting (software) engineering as a profession to having utter disgust with most of what I saw and did.”

For one person, the pressure to produce faster was part of a much wider change in the culture of Silicon Valley. They describe watching a workplace they had once taken pride in become increasingly difficult to recognise, as layoffs, heavier workloads and the expectation to keep producing reshaped what it felt like to work there.

“I cried at my desk a lot those days. I remember going from over 200 people in the company to 30. It was brutal.”

“I went from taking pride in my work to hoping I made it through the day with some dignity intact.”

That sense of alienation appeared in another interview, too, but this time around something more specific: whether workers felt able to admit what they actually thought about AI.

“No one knew how their peers felt because of how taboo it became to talk about anything other than how productive you were becoming due to your growing use of the technology.”

They called it “a fetish for overproduction.”

Forget Big Brother. In Silicon Valley, the C-suite is watching.

And if admitting reservations about that technology becomes taboo, the problem goes beyond workplace morale. It means the people closest to the products may become less willing to say when something is wrong with them.

The concern is therefore not only what workers are being asked to build, but whether they still feel able to question it.

So if the pressure is coming from the top, what is driving it?

The answer is to follow the money.

Follow the Money

“My scepticism stems from the glow of capitalism and greed, which shines so bright in this country.”

There is a reason that line stayed with me. By this point in the conversations, the pressure to build faster and produce more no longer seemed like an isolated problem with individual managers or companies. The people I spoke to repeatedly came back to the same thing: money.

So why do people continue to work under these conditions?

One person I spoke to was particularly candid about the difficulty of resisting an industry that can pay extremely well.

“I don’t know if anyone else would willingly turn down a (very) large paycheck for what seems like a fruitless endeavour, without the promise of a safety net. Everyone enjoys having a roof over their heads and food on their plates.”

That is perhaps the less glamorous reality behind Silicon Valley’s enormous salaries. Walking away is easier to imagine when you are looking at the industry from the outside than it is when your mortgage, your career and your financial security are tied to it.

But the financial incentive does not stop at the worker. Companies have their own reasons to keep moving. One source described a growing desperation to “outproduce and outmarket all the competition”, while another said the C-suite at their company had become focused on “marketing, new product, and efficiency.”

And then there is the question of what all that productivity is actually supposed to achieve.

For one former worker, the answer became increasingly clear: replacing skilled labour.

“The end goal of this technology was to replace workers. Not making bad jobs obsolete, but making skilled jobs no longer require skilled workers.”

Then came the phrase that perhaps best captures their concern:

“Giving the moneyed access to skill while removing the skilled’s access to money.”

It is a very different interpretation of the promise that AI will make workers more productive.

If a programmer can use AI to do in an hour what once took them three, the obvious question is who gets the benefit of those two hours. The worker could gain more time, better pay or a less exhausting job. The company could instead decide that it needs fewer programmers.

That distinction sits at the heart of the concerns I heard from inside the industry. Productivity is not simply a technical achievement. It has economic value, and whoever controls that value gets to decide what increased productivity means for everyone else.

The same incentives help explain the race to adopt the technology in the first place. If your competitors are moving towards AI, standing still can look like falling behind. If investors expect AI to transform productivity, companies have an incentive to demonstrate that they are doing something with it. If everyone is racing, being the company that decides to slow down suddenly becomes a much harder proposition, especially in a culture arguably obsessed with profit.

Can we trust corporations to voluntarily slam the brakes on dangerous or unethical technology when the precedent being set is a relentless pursuit of profit?

Because slowing down has a cost. So does deciding that a product needs more testing, that a particular use is too risky, or that replacing workers is not worth the social consequences.

The people I spoke to are not simply describing an industry that wants to make money. They are describing an industry where enormous financial incentives are attached to getting ahead, getting bigger and, importantly, being the fastest to do so. A so-called AI arms race.

When that is the environment in which AI is being built, the question of who benefits becomes impossible to separate from the question of what gets built in the first place.

And that brings us to the question the industry cannot answer through productivity figures or market valuations: should everything that can be built actually be built?

Should We Build It? AI Ethics

Just because we can build something, should we?

“When we toss away asking these questions, we chip away at the quality of our profession.”

For the people who build technology, this is supposed to be part of the job. What could go wrong? How could something be misused? What happens when a product leaves the hands of its creators and ends up in the hands of millions of users?

One person believes those questions are increasingly being neglected.

“The newer generation of engineers, project managers, and designers really do not care about building within the confines of the law or ethics. Most aren’t even aware of the legal implications of what they’re building.”

It is a serious accusation, particularly from someone who has spent years working inside the industry. But their concern is not simply about whether individual engineers are making mistakes. It is about whether the culture of building technology has changed so much that asking whether something should be built is becoming secondary to working out how quickly it can be built.

They point to the people making decisions at the top of companies, describing the C-suite as interested primarily in “marketing, new product, and efficiency.”

And that leaves a question hanging over the people actually writing the code: who is responsible for stopping something when the people building it believe it should not go any further?

That question becomes particularly striking when the people inside the industry are themselves deeply uncomfortable with the technology.

“My Mac runs El Capitan (circa 2015). My iPhone runs on OS 17.6. The moment AI came out, I intentionally shut my tech down. I may sound paranoid, but I don’t think I am.”

There is something almost absurd about a technology worker deliberately keeping their own devices years behind to avoid AI. But beneath the irony is a much more serious point. This is someone who has spent years close to the technology and has come away sufficiently concerned that they do not want it in their own life.

That does not make their fears automatically correct. It does, however, make them difficult to dismiss as the anxieties of someone who simply does not understand AI.

This is the view from inside: people who know how these systems are built, who have worked alongside the companies developing them, and who are still asking whether some things should be built simply because they can be.

For the workers who raised these concerns with me, ethics is not an abstract question to be debated after the product is finished. It is supposed to be part of the building process itself.

And when those questions are pushed aside, they believe something fundamental is lost: not just safeguards for the eventual user, but the professional responsibility of the people creating the technology.

So, What Does This Mean for Us?

So what does any of this actually mean for the rest of us?

After speaking to people who have spent years inside the companies building AI, I am not sure the answer is that we should all throw our phones in the sea and start communicating exclusively by carrier pigeon. Nor is it that AI is inevitably going to destroy everything we know and love.

What I do think it means is that we should pay considerably more attention to the people building it.

Because once you step inside the industry, the story becomes more complicated than the one we are usually sold. There are workers who believe AI could be useful, workers who are deeply uncomfortable with what they are being asked to build, workers who have watched their workplaces transformed by the technology and workers who have ultimately decided they cannot be part of it anymore.

They are not all making the same argument. But there is a common thread running through their experiences: the pursuit of profit is repeatedly being placed above ethics, quality, worker security and sustainable development.

The engineer can question the product, but the pressure to get it out the door remains. The worker can have reservations, but productivity is still the benchmark. Someone can raise concerns about the consequences of AI, but companies still have shareholders to satisfy and competitors to outrun.

And eventually, those decisions leave the office.

They reach the worker whose job is changed or disappears. The person using an AI system without knowing how it was developed or what safeguards exist around it. The creative worker wondering whether their work is being replaced. The communities living with the infrastructure required to keep these systems running. The public, increasingly being asked to trust technology that was designed and deployed long before most of us had any say in whether we wanted it.

So what does that leave us with?

One of the people I spoke to had an answer: “vote with our dollars.” They described consumer spending as “the only vote they actually listen to.”

It is an uncomfortable conclusion because it puts some responsibility back on us. We may not have a seat in the boardroom deciding how quickly an AI product is released, but we do have some say over which products we use, which companies we support and how much of this technology we allow into our own lives.

That does not mean individual consumers can solve a problem created by some of the most powerful companies in the world. But it does mean we are not entirely powerless either.

And perhaps that is the most important thing I took from going inside the industry. AI is often presented as something happening to us, as though the technology itself is an unstoppable force arriving from somewhere in the future. But behind it are people making decisions now: what gets built, how quickly it gets released, what risks are acceptable and where the line is drawn.

Those decisions are being made inside companies with enormous financial incentives to keep moving.

So perhaps the question we should be asking is not simply whether AI is good or bad, or even whether AI itself can be trusted.

Can we trust an industry to build technology responsibly when its most powerful incentive is to make money from building it as quickly as possible?

The people inside have given us a glimpse of what that process actually looks like. Now the question is what we do with what they have told us.