*This episode was first broadcast on March 25, 2026.
Karen Hao, a tech journalist with a background as an engineer in Silicon Valley, has become a vocal opponent of the major AI corporations and their aggressive pursuit of artificial general intelligence, prioritizing “scale at all costs.” She asserts that these entities have established a global system that collects data and intellectual property, infringes on human rights, and incurs significant environmental damage.
“At its core, we must disconnect AI from empire. We desire the advantages of AI, but we cannot accept them at the expense of our democracy,” Hao expressed during a public address at the University of Toronto’s Schwartz Reisman Institute for Technology earlier this March.
“People should fiercely defend against any attempts to take that away,” she remarked during her interview with host Nahlah Ayed.
In her book Empire of AI: Dreams and Nightmares in Sam Altman’s OpenAI, Hao explores the worldwide consequences of Big AI, discussing the ramifications of its swift implementation, the decline of democratic principles, and the necessity for a responsible rethinking and redesign of AI systems that are more purpose-driven in the future.
As she encouraged her audience during her lecture, “Let’s envision new forms of AI that can deliver benefits without incurring extraordinary costs.”
Here is a passage from Karen Hao’s speech:
We must cease viewing [AI] companies as simply businesses offering products and services. These are emerging empires consolidating an unprecedented amount of economic and political influence, altering our planet, redefining geopolitics, and disrupting our educational frameworks and future professions.
Why do I refer to them as “empires?” The AI empires function in much the same way as historical empires did. In my book, I highlight four parallels.
To begin with, they claim ownership of resources that do not belong to them, including individuals’ data and the intellectual property of creators such as artists and writers.
Secondly, they exploit a vast amount of labor. This includes not only the workers who contribute significantly to wealth generation for these companies without receiving comparable rewards but also those who are displaced by automation once these technologies are deployed. This reflects a deliberate political decision made by these firms to transform their AI into systems that automate labor.
Thirdly, they monopolize knowledge creation.
Over the past decade, the AI industry has become the leading employer and sponsor of AI research. Consequently, they can dictate the research agenda while also censoring and controlling inconvenient truths.
Thus, the public’s understanding of the capabilities and limitations of these AI systems is influenced by what the empire chooses to disclose. It’s comparable to a scenario where most climate researchers are funded by fossil fuel industries; the public would not receive an accurate depiction of the climate crisis.
Lastly, these corporations rationalize their actions with a moral and existential justification. They position themselves as the benevolent empire on a mission to civilize and modernize humanity. They claim that granting them access to all data, resources, and labor will lead to a utopian existence. Conversely, if they fall to a malevolent empire, humanity faces a descent into chaos.
This raises the question: is this truly necessary? Do we require empires to advance AI? Must we rely on empires to benefit from this technology? To explore this question, let’s use an analogy.
One significant challenge in discussing AI today is the ambiguity surrounding the term artificial intelligence. It’s akin to the word transportation, which could refer to anything from a bicycle to a rocket. Clearly, these represent different modes of transportation, designed for various purposes and with distinct cost-benefit considerations. The same applies to AI; it encompasses a broad range of technologies.
Thus, when we inquire about how to gain benefits from AI, we must be specific and identify which AI technologies we desire more of and which ones we should minimize. Additionally, we need to consider how to redesign, enhance, and develop new AI forms where the advantages outweigh the drawbacks.
I would argue that the prevalent AI systems capturing headlines and public imagination today, particularly large-scale general-purpose systems like ChatGPT, exemplify the least favorable trade-offs in our array of existing AI technologies. This is the version of AI that Silicon Valley encourages us to adopt, enabling them to expand their empires while imposing significant costs on many segments of society.
If we genuinely wish for AI to yield broader benefits, an urgent shift away from this approach is necessary.
*Excerpt modified for clarity and brevity. This episode was produced by Lisa Godfrey.
Download the IDEAS podcast to listen to this episode.
Tag: rewrite-pending

