Karen Hao is the author of the bestseller Empire of AI – Inside the Reckless Race for Total Domination, which was published a year ago. She started out studying mechanical engineering at MIT then briefly joined a tech startup before switching to journalism. On Friday, she spoke to co-founder Carole Cadwalladr at London’s Conway Hall for an event organised by How To Academy in collaboration with the Nerve. This is an edited version of the conversation.
Carole Cadwalladr: In 2019, you wrote this seminal article in MIT Tech Review – it was the first time a reporter had got inside OpenAI, and at that time it was very secretive, quite mysterious. It had very high-minded ideals. It was a non-profit. That was the beginning of this journey, in a way, wasn't it?
Karen Hao: I didn't think that I would end up writing a book about OpenAI, because I didn't think that it would become the avatar for this entire AI era that we're in. OpenAI was in the early days of its journey, it was making a little bit of a mark. It was just serious enough for a profile, but not so serious to give to the senior AI reporter, and that's why it went to me. I was very early in my career, and it was actually my editor at MIT Tech Review that even suggested that I do a profile. And so I embedded within the organisation for three days.
OpenAI was portraying itself as really open, transparent, welcoming. I had this impression that I was coming to an organisation that was going to welcome me with open arms. But then once I got there, that's when things started to feel a little bit off, because they were really controlling about what I was allowed to see, who I was allowed to talk to, in a way that is much more normal for a company, but not for a non-profit that says that it was supposed to be fully transparent. So that's when I started asking more questions.
The very first meeting was with Greg Brockman, at the time chief technology officer, now president, and Ilya Sutskever, who was the chief scientist, and either the first or the second question that I asked was: why spend billions of dollars on this particular problem? You know, you could spend billions of dollars on many things that would be good for society. And Brockman paused for around 12 seconds, and I was like, that's not good. That was a really softball question, and the most senior leaders leading the day-to-day operations aren't able to articulate what they're actually doing at this organisation. And that's when I started asking a lot more questions.
What was your headline finding?
I suppose it was the idea that OpenAI was ultimately not portraying itself truthfully, and it had created this public narrative very detached from what it was actually doing, to accrue an enormous amount of public support and an enormous amount of capital. I think I was honestly quite nice to them. But it was beyond what they were willing to swallow, and I was blacklisted from ever interviewing people at the organisation again. I later found out that [chief executive] Sam Altman sent out an email to the rest of the company that was basically like, “yeah, this is bad, and it's going to affect our fundraising, but let's just put our heads down and do good work and prove this reporter wrong, instead of creating a kind of public fight that gives her profile more oxygen”.
Actually, being cut off by the company was probably the best thing that could ever happen to you, because as we know, one of the biggest problems in the coverage of these tech companies is that they cultivate journalists and drip-feed them information, basically in exchange for propagating the narratives that they want.
I was young in my career, and I didn't really understand the game of access journalism, so I did what I thought I was supposed to do, which is just to tell exactly what I saw and hold power to account. And I was really alarmed when I got cut off from OpenAI, in that I thought “did I misunderstand what I was supposed to do?” But I had fantastic mentors that were like, “you're not supposed to play the access game: they close one door, there will always be another one open, you just have to find it”. So I had to learn early on to do my coverage without that front-door access.
How prevalent would you say that style of access journalism still is?
Hugely. They are preying on the vulnerability of the fact that journalism as a business model is really failing, and a lot of publications do rely on these scoops to drive traffic, to drive audience, drive subscriptions.
In your book, you interview over 250 people to build an incredible portrait of Sam Altman, who his board eventually, when they tried to get rid of him, called not consistently candid. If Sam Altman was a bit more of an upright guy, would that be better?
It's important to understand the character of these individuals, in part to bring them back down to earth, because so many people think that this technology is neutral, that it doesn't actually come from human decision-making. One of the things that I wanted to do by profiling the individuals was to demonstrate the degree to which their decision-making can often be driven by extremely petty reasons – rivalries, ego, personal matters – that create a lot of anxiety and pressure for them, and also to just demonstrate that they're they're flawed people. Everyone's a flawed person, but to show these people are like any other person. The problem is they just have accrued an enormous amount of power, so that is the extent to which character does matter.
But I find that sometimes people take this to the opposite extreme, where they start obsessing with this question of is this a good person or a bad person, and if it's a bad person, should we just swap them out for a good person, and does that solve the problem, and that's also not the right way to think about it. The fundamental issue is a structural one. The structural problem is that these people, whoever they are, get to occupy positions in which they make decisions for billions of people around the world, and those billions of people don't have formal mechanisms by which to provide feedback and hold those people accountable.
AI is a political project, and the central feature of this project is to take agency away from people to make these entities the monopolies on decision-making power
Can we talk about the fightback? Have you seen the viral videos of the commencement speeches in the US? My favourite was seeing Eric Schmidt, latterly of Google, getting booed from the stage. So the kids are all right?
The younger generations always get it faster than everyone else when it comes to new technologies. I mean, my generation was the social-media-era generation, the millennials, and we were the ones that started asking the questions. Is this actually good for our brains, is this good for our mental health, should we be detoxing?
That's just the same thing that's happening now with gen Z, gen alpha, where they're seeing the corrosiveness, whether it's to the environment, or to their education and their critical thinking, or to their future careers, because they are the generation that's going to be hit the hardest by all of this stuff. We're setting ourselves up for a divisive intergenerational battle here. Because people who are much older, they're invested in the stock market, or have their pensions in there.
What about other forms of resistance? The other thing which has been happening in the US is this amazing local mobilisation against data centres.
This has been my favourite development since the publication of my book: that there have been so many different communities around the world, but in the US in particular, that have suddenly realised that they actually can do something about the ways that the AI industry is expanding, and one of the biggest mobilisations has been around the data centres. In 2025 in the US, data centre protests have successfully stalled over $100bn worth of projects. I was recently in New York at an event where I met a bunch of activists from different communities and there were a couple things that were interesting. One was that it really cuts across different coalitions in society. The environmental groups are coming out, the working class, people that are worried about noise pollution, or they're just simply pissed off at the tech oligarchs.
What was remarkable is these activists said “we fight not just for our community, we fight for every community”. They are motivated by this deep sense of purpose that extends well beyond their own land. It's such a beautiful idea of collective action and acting on behalf of a much broader coalition of society that is now extending to other modes of resistance, not just data centres.
There have been workers that are unionising, organising across different sectors, in the AI industry itself. Workers that are the most exploited and doing data annotation and content moderation work – mobilising, organising, elevating their plight to international media attention to shore up the labour practices within the industry. Also, tech workers within the companies themselves, that are getting paid really high compensation packages, are now also beginning to organise, delivering open letters.
Empires don't just claim all these resources, they also need these resources, and when they are withheld from them, it begins to make their very foundations shaky
There was this Amazon open letter that came out several months ago. You know, tech workers, tech employees, are often quite vulnerable – even if they're paid really well, they're deeply surveilled. Many of them are in situations where their healthcare depends on these companies, their immigration status depends on these companies, and in a moment when the industry is aggressively trying to surveil these workers, they were still able to put out an open letter with over 1,000 signatories demanding that their leaders at Amazon stop engaging in a deeply corrosive approach to AI development, and to articulate and try and use their collective worker power to scrutinise their own organisation.
The prevailing narrative is that this is going to happen whether we like it or not – these incredibly powerful companies, you can't stop progress. It turns out it's not progress, and you can stop it. AI, as it is currently conceived by Silicon Valley, is a political project, and the central feature of this political project is to take agency away from people to ultimately make these entities, the empires, the monopolies on decision-making power in the world, if they can convince everyone that this is inevitable. “Don't do anything about it,” you know, “just lie back and let it wash over you”. They have achieved their goal, which is to make sure that people don't feel any ownership over their life, their data, their land, their resources, and their future any more.
What these acts of resistance are is the reclaiming of that agency, the recognition that, in fact, wait a minute, we can do something about it, and all of the resources that you need are in fact our resources. The data centre protests are saying “this is our land”, and the workers are saying “this is our labour, this is our economic opportunity that you are trying to take away from us”.
What's amazing is the empire metaphor continues to bear fruit, because empires don't just claim all these resources, they also need these resources, and when this land, this labour, is withheld from them, even a small fraction of it, it begins to make their very foundations shaky.
So recently OpenAI had to shutter its video generation tool, Sora, which initially they announced as a product that was the second best product since ChatGPT. Why did they shutter it? If you look at all the reasons, every single one was shaped by grassroots action. One reason they were incredibly constrained on computing resources. A lot of these data centre protests have stalled OpenAI projects.
The second reason they had is they're facing an increasingly uncertain financial future. They're setting up for an IPO [stock market flotation], and they're not as sure any more about whether or not they'll be able to continue amassing gobs and gobs of capital. This goes to the way that Wall Street investors price and value these companies, and Wall Street investors have been looking at the broad backlash that is happening against AI and have gotten more and more nervous that these companies won't be able to achieve their goals, and therefore started devaluing these companies a little bit.
And the third reason is that the product wasn't popular, the usage was flatlining, and so OpenAI wouldn't have been able to monetise it anyway – and so that's another form of collective consumer action.
It's all grassroots action. The literal resistance actions that people are engaging in is chipping away at the pillars that uphold the empire, and the empire is beginning to crumble.
Empire of AI: Inside the reckless race for total domination is published in Penguin paperback, £12.99
For information on future How to Academy events