Feminist AI Scientist: What would a feminist ChatGPT look like?

taz: Ms. Kubes, what was your motivation for the book “Code & Care”?

Tanja Kubes: The political moment. The most powerful technology of our time lies in the hands of a few men who openly collude with authoritarian forces. The USA and China are engaged in a race in which security and equality are seen as brakes. The federal government's digital summit shows how urgent the questions in my book are: by 2030, four out of five large companies and nine out of ten federal authorities are expected to use AI, and the Digital Minister would like “more freedom” in regulation. Anyone who introduces AI so widely without asking who built it and who it disadvantages is digitizing the injustices of yesterday. That's why my book asks: whose future is being built here?

is a doctor and expert in sociotechnical systems and has been researching, teaching and advising on AI and robotics for over ten years. It combines power-critical sociology with more-than-human ethnology.

taz: You have already developed the “Feminist AI Framework” (Faif) in 2025, a kind of manual on how AI could be designed in a feminist way.

Kubes: Yes, and it goes beyond previous approaches, including feminist ones. They usually ask who is being discriminated against. Faif also asks what is excluded: knowledge systems, relationships, ecosystems. To do this, I am combining six perspectives for the first time, from feminist criticism of power to indigenous knowledge systems to the philosophy of the body. From this I developed a three-stage FAIF certification that sharpens the EU AI Act where it remains vague.

taz: In your book you ask speculative questions at the end of each chapter, such as: “What if feminist peace researchers had financed the first AI systems?”

Kubes: AI grew up with Pentagon money, control and surveillance are still in its DNA today. With feminist funding, the first machines might not have played chess, but translated sign language or helped end world hunger. Questions like this show that nothing has to be the way it is. Because imagination is the first place where power becomes visible. If you can tell different stories, you can build different futures.

taz: In the intro you explain that you consciously decided to write the book in simple language and not as a scientific work. Why?

Kubes: Because AI affects us all, whether we like it or not. She decides on applications, loans and finding a partner. To ask who is building something like this and for whom, you don't need to be able to read code. I didn't want to leave my knowledge from over ten years of research in the academic echo chamber, but rather make it available to everyone.

taz: You write that care runs as a common thread through your book. How can one understand this?

Kubes: Care is the alternative to an industry that exploits people, data and nature. Firstly, it means making invisible work visible: without people who label and filter data, no AI works, but it disappears behind the “intelligent machine” like housework behind the functioning household. Second, it says relationships over efficiency. And thirdly, care does not end with people, but includes rivers, soils and animals. A chatbot's response comes in three seconds, as if by magic. Behind it, however, are cobalt mines in the Congo where children also work. Data centers like Google's in Oregon, which uses over a billion liters of water in a year. People in Nairobi sorting through traumatizing content for less than two dollars an hour. Wealth flows upwards, costs downwards.

Kubes: Yes. The Global South supplies raw materials, cheap labor and data, the Global North collects profits and writes the rules. Then there is the language: AI primarily learns from a few world languages ​​such as English and Chinese. Thousands of smaller languages ​​are barely used, and with them the knowledge and worldviews of billions of people disappear from the systems.

Tanja Kubes: “Code & Care: Making the world fairer with feminist AI”. Publisher Barbara Budrich 2026, 24 euros

taz: Many people argue that technology and AI are neutral because they are based on data. What do you say to that?

Kubes: Then I invite you to my seminar. There, students enter “STEM professor” into image generators. Beforehand, almost everyone was convinced that AI was neutral. Then all they see is older white men in lab coats. At “STEM professors having a party” sexy robots with female shapes serve them pizza and drinks. This is sexism. At “STEM professor in Africa” the professor becomes a primary school teacher who teaches children in front of a mud hut. This is racism. The image generator doesn't make a mistake, it has a point of view. Data is a thing of the past. And the idea of ​​an “objective machine” is one of the most powerful illusions of our time.

taz: What would a feminist ChatGPT look like?

Kubes: Above all, a feminist chatbot would not be a purely capitalist product. Today's chatbots are designed to ensure that we talk to them for as long as possible and feed in as much data as possible. That's why they constantly confirm us and prefer to invent sources rather than admit a gap in their knowledge. A feminist chatbot would not be optimized based on how long it spends, but rather based on whether it really helps us. He would ask: whose voice is missing from the data? He would reveal his point of view and admit when he doesn't know something. It didn't belong to a billionaire, it belonged to the community. And he wouldn't be a substitute friend, but would send us back to other people.

taz: In your book you also criticize the media coverage of AI – what exactly?

Kubes: That she often retells the fairy tales of Silicon Valley unchecked and far too rarely asks about the power structures. We are experiencing an incredible power asymmetry: a handful of corporations own the models, the data centers, the platforms and our data. This power increases with every new system. At the same time, fear is being fueled that AI could take over the world or destroy humanity. That's distracting. Anyone who mystifies AI makes it uncriticizable. The real danger is not superintelligence, but ordinary AI in the hands of a few.