Overthrow of the AI empire

While the technocratic-authoritarian race to develop advanced artificial intelligence technologies continues at full speed, the influence of Silicon Valley on regulatory agendas already rivals – and even surpasses – that of many states. However, among the ambitions of global superpowers and AI giants, Europe has the opportunity to seek something better. Interview with the award-winning journalist and author, Karen Hao.

Alessio Giussani: Several metaphors have been used to describe the current model of AI development – for example, a parrot or a mirror. Why did you choose the empire metaphor specifically?

Karen Hao: I believe these metaphors are all extremely useful. I focus on the "empire" because I am not only examining the technical tool itself and what happens when it is applied in society, but also the production of the technology, as well as the actors who amass extraordinary economic and political power to produce it.

I call this system an empire for four main reasons. First, these companies require resources that are not theirs: individuals' data, as well as the work of writers and creators. Second, they exploit labor to an extraordinary degree, affecting both workers exploited in the manufacturing process and supply chains, and those whose labor rights are undermined during the implementation of the technology.

The third aspect is information control. A key feature of empire expansion is the ability to filter all knowledge production and information through their imperial goals, and this is exactly what AI companies do. They control and censor science by financially supporting most AI research and building information technologies that they aim to centralize into a portal through which everyone else connects to the world, including education.

Finally, they cloak all this in a civilizational mission. They claim that their work ultimately aims to develop technology that elevates humanity to a utopian state, but if the "bad guys" – usually China – develop it before us, we will instead fall into a dystopian state. This is a kind of religious narrative, which was also common among ancient empires.

The AI empire, as you describe it, is not limited to a single company; it is a system, a logic. What is behind the attention directed at OpenAI and its CEO, Sam Altman?

One answer is that people need stories to help concretize what would otherwise remain an abstract idea. But OpenAI is especially interesting because it has made all the major decisions that have changed industry norms and steered it toward empire-building. OpenAI was the one that decided to expand at all costs, working with large language models, and to create a chatbot-like interface for them. Almost everything the public associates with AI today was first decided within OpenAI's walls.

If we tell this story, it becomes much clearer that AI is not inevitable, because we realize that real people made those decisions based on their values, worldview, and personal revenge – whatever actually shaped the trajectory of AI development. Sam Altman's ideologies are not just abstractly embedded in the model. We often hear the cliché that models reflect their creators, but I wanted to show how this actually happens. It involves companies, people, and the economic and ideological environment they are part of.

As these technologies become more efficient, gaining control over their application becomes an existential issue. We saw Big Tech bow before Trump during his second inauguration, but we also see CEOs of AI companies increasingly opposing the government. Who will have the final say – the state or private companies?

The ongoing battle is over who will make the decisions. I see the relationship between Washington and Silicon Valley as a convenience alliance, but a very tense one. They represent two different imperial models, each with its own imperial ambitions. Both seek to expand and literally want to take over more territory from other countries, but the Trump administration views AI companies as tools for the state's imperial ambitions. Meanwhile, Silicon Valley sees the Trump administration as a tool for its own imperial pursuits, but in reality, they are not on the same wavelength, because both want to be the ultimate authority at the top.

There have been moments when the US government proved it was still the boss – for example, when Trump and Musk clashed, Musk was fired. But when the Department of Defense clashed with Anthropic, the company proved it was the stronger party: Defense Secretary Pete Hegseth proposed the "nuclear option," threatening to declare the company a supply chain risk – as a last resort to force Anthropic into compliance – but even then, the company refused to yield. Embarrassingly, the Department of Defense became deeply dependent on a private actor that refuses to cooperate under government conditions, and the government cannot do anything about it.

Both imperial actors must be held accountable, and this accountability will not come from either of them; it must come from the people. From bottom-up, through grassroots organizing, which we are beginning to see flourish worldwide, including in the United States. No matter how impenetrable this convenience alliance seems, it is actually extremely fragile, and there are opportunities to accelerate its downfall by exerting greater pressure on its weak points.

Instead of viewing artificial intelligence as a global competition between good and evil, we should think more about what kind of future we want for society.

The current state of AI development appears as a competitive race towards a low point between rival companies and global superpowers. What role did the geopolitical competition with China play in bringing AI to its current difficult situation?

China is a card that Silicon Valley has been playing for a decade, essentially since they expelled China from the market. But if we look at the previous outcomes of this argument, we see that in reality, it was always the opposite of what Silicon Valley predicted.

They claimed that the lack of regulation in Silicon Valley – and China's export restrictions and regulation – would allow American technology to dominate worldwide. Yet TikTok is the most popular social media platform, and Chinese open-source AI models are becoming increasingly popular globally, including in the US. They also promised that Silicon Valley's dominance would have a liberalizing effect on the world, but that did not happen; instead, their influence has been deeply illiberal. There has been a significant erosion of democratic institutions and a rapid acceleration of democratic decline, facilitated by Silicon Valley platforms and AI technologies.

Instead of viewing AI as a global competition between good and evil, we should think more about what kind of society we want to build, which AI technologies would support this vision, and what resources we need to realize it. We should not compete in a race dictated by Silicon Valley, but rather in the realm of ideas and values we want to uphold. And I believe Europe has an opportunity here to show the world a third way, different from the US and China, establishing new rules.

The European Commission recently adopted a technological sovereignty package as part of its effort to chart its own technological path. How can we articulate this future so that we do not end up with a surveillance-based, oligarchic business model in Europe?

The guiding star Europe must follow is not sovereignty for its own sake, but the goal of achieving that sovereignty. Europe – or at least a large part of it – is one of the last bastions of democracy in the world. It supports human rights and freedoms, and it needs technological sovereignty to promote these values. As long as it remains heavily dependent on systems developed in an authoritarian manner and aimed at undermining European societies, it cannot uphold its values.

The concept of sovereignty tends to blur if we start interpreting it as an end in itself. It must be a means to achieve a goal.

What coalitions, interest groups, or conflict arenas should we start from to build this alternative future?

First, we need to understand AI governance in a broader sense. For example, strengthening workers' rights and protecting civil society are part of AI governance. Students, teachers, and healthcare workers must also be at the table to help shape and question how AI is developed and applied in society. Education is key, both as a sector and as a coalition we need to involve, as well as for raising social awareness about AI.

Another important aspect is funding for AI researchers and entrepreneurs to explore the technical details of their development and its applications. But we cannot simply give them money without giving a voice to the rest of society; otherwise, we risk repeating the problems we face in the US, where technocrats or technoconservatives make decisions.

We are not starting from scratch. Much work has already been done to define shared goals that society should strive toward. For example, I think of the Sustainable Development Goals. This can be a starting point for figuring out how to finance innovation in AI.

The narratives around current AI technologies are highly polarized. On one side are those who say these technologies are nothing particularly new and that the bubble will soon burst. On the other side are those who believe AI will transform our lives, revolutionize supply chains, and make us all redundant. What do you think about these different positions?

Indeed, many tend to exaggerate the benefits without considering the costs of the technologies, or they overemphasize the costs without seeing the benefits. But I believe there is a significant proportion who agree that we want to enjoy AI's benefits, but not at the expense of our dignified way of life – which includes good healthcare, economic opportunities, a healthy environment, and a thriving planet. This is the most balanced approach.

Artificial intelligence (AI) is a broad term covering many different technologies, so we need to carefully consider which to invest in and which new technologies to develop, because each involves different trade-offs; some align with the future we want to build, others do not. We must choose wisely.

The term artificial intelligence (AI) refers to many different technologies, so we need to carefully consider which to invest in and which new technologies to develop.

What is your relationship with existing AI tools?

I do not use them for three main reasons. First, because of my ethical stance. Second, because of my stance on privacy: I examine these companies, and ultimately they are some of the most effective surveillance technologies ever invented. Third, because these tools are not really useful for the specific skill set I need to do my work.

I am not saying everyone should do the same. I have spoken with many people who clearly benefit from these tools, or for whom their professional role makes their use mandatory. The better approach to the current problem is to do what we did in other industries in the past, like fashion. When we realized that most of our clothes were made under terrible conditions, we didn't just get rid of the clothes. It took a range of consumer and worker organizations, government regulations, and public backlash to rein in the industry. Of course, the fashion industry is still far from perfect, but today there are alternatives, and consumers can vote with their wallets for supply chains and corporate practices they support.

The same must be done with AI. Currently, there are not enough alternatives for users, and to change that, we need more worker organizations and greater public resistance. Those who can abstain from using these technologies must boycott them, and governments must intervene to help create new markets for alternatives.

Consumers have significant influence, but it is not the only one. As individuals, workers, educators, and citizens, we must consider how to leverage these roles to improve AI development.