DIGITAL LIFE

Understanding how feminist AI in Latin America works

Online spaces perpetuate stereotypes about who creates them and the data that is inserted into them — currently, predominantly men. This global phenomenon has been driving the construction of technological alternatives, such as the Feminist AI Network of Latin America and the Caribbean.

The technological literature is full of examples of gender bias. Image recognition systems have difficulty accurately identifying women, especially Black women, which has already resulted in misidentifications with serious consequences for law enforcement.

Voice assistants have long used exclusively female voices, reinforcing the stereotype that women are more suited to service roles.

In image generation, AIs often associate the term "CEO" with a man, while a search for "assistant" returns images of women.

— Artificial intelligence feeds on data that is not neutral: it reflects societies marked by historical inequalities and power relations. If a company wants to achieve fair results, it needs to analyze datasets, verify their representativeness, and actively intervene when this is not the case. Equity doesn't just appear on its own: it needs to be designed — Ivana Bartoletti, an international expert in AI governance and author of a Council of Europe study on artificial intelligence and gender, told the ANSA news agency.

For Bartoletti, the recent case of Grok — Elon Musk's AI that allowed the generation of fake images of naked women and minors, a function that was later discontinued — “shows what happens when the safety and rights of women are not considered in the design of systems.”

— If there are tools to undress women, they will be used. Deepfake nudes are a form of humiliation and control. The implicit message is dangerous: you are online, therefore you deserve this. This is how many women are silenced and abandon the digital space — she explained.

It is in this context that technological alternatives emerge to rethink artificial intelligence and transform it into a space of struggle and shared power.

Feminist AIs...In Latin America and the Caribbean, for example, the Feminist AI Network emerged, supporting dozens of projects focused on transparency and public policies. Tools like AymurAI, Arvage AI, and SofIA apply a gender perspective to legal analysis and expose the biases and discrimination inherent in algorithms.

Afrofeminism has also been reclaiming artificial intelligence as a space for self-determination, with assistants like AfroféminasGPT, trained based on the knowledge and voices of Black people.

— They demonstrate that we can organize ourselves to use AI for the benefit of all, share data collectively, and develop solutions centered on real needs. But the key remains power. The feminist issue in AI is a question of power: women need to have more power. Not on the margins, but at the top of companies and in the spaces where technological policies are decided. We need diversity in decision-making environments, not just among programmers. Artificial intelligence is not just technology, it's a choice about how we want to transform society — concluded Ivana Bartoletti.

by La Nacion — Buenos Aires