The rapid rise of artificial intelligence is prompting increased scrutiny of its often-hidden costs, extending beyond technological advancements too encompass human rights, environmental sustainability, and geopolitical concerns. As the AI sector-currently valued in the hundreds of billions-faces mounting pressure for responsible development, investigations reveal a complex global supply chain reliant on ethically challenging labor practices and resource extraction [[1]].This report details these concerns, alongside emerging efforts to build a more equitable and transparent AI ecosystem.
The rapid advancement of artificial intelligence is increasingly scrutinized for its hidden costs, extending beyond technological innovation to encompass significant human, environmental, and political implications. Concerns are mounting over the dominance of U.S.-based companies in the AI landscape, prompting calls for a more responsible and equitable approach to development and deployment. The growing AI sector, currently valued at hundreds of billions of dollars globally, is facing increased pressure to address these ethical and logistical challenges.
The resources required to power AI are far from virtual, according to recent reporting. The sourcing of raw materials, often from regions with challenging and dangerous working conditions – such as the Democratic Republic of Congo – is a critical component of the AI supply chain. This reliance on resource extraction raises questions about sustainability and labor practices.
Beyond materials, the training of AI models relies heavily on human labor, often performed under precarious conditions. A significant portion of the work involved in refining and improving language models, particularly those designed for French speakers, is outsourced to workers in Madagascar. Despite the perception of AI as highly automated, these models are, in reality, heavily supported by human intelligence.
>> Relire : Derrière les prouesses de l’IA, l’exploitation de travailleurs invisibles and En Inde, des petites mains gavent l’intelligence artificielle de données
The vast datasets used to train AI models are often compiled without obtaining consent from the original creators of the text and images. This practice raises concerns about copyright infringement and intellectual property rights, issues that are now attracting regulatory scrutiny.
>> Lire aussi : Google visé par une enquête de Bruxelles sur l’utilisation de données pour l’IA
>> Ecouter à ce sujet :
While new digital tools offer significant potential, there’s a growing concern about the concentration of power within a handful of major technology companies. “It wasn’t necessary to create a digital world where we use products from only four or five large companies,” one analyst noted. This lack of diversity raises questions about innovation and market competition.
A shift towards a more balanced AI ecosystem requires the development of “a multitude of different models,” mirroring the diversity found in the media landscape. This plurality is seen as essential for fostering democratic principles and providing consumers with genuine choice.
“To build more responsible and respectful digital worlds and AI, we must move away from the idea of creating a ‘white knight’ to challenge the GAFAM companies,” – Google, Apple, Meta (formerly Facebook), and Microsoft – experts suggest. The focus should be on fostering a broader range of solutions rather than seeking a single dominant alternative.
Growing opposition to the construction of “data centers,” the massive facilities required to store and process AI data, is another sign of increasing scrutiny. Simultaneously, initiatives like the Tournesol project at the Swiss Federal Institute of Technology in Lausanne (EPFL) are exploring new types of recommendation algorithms that prioritize collective intelligence over the approaches used by Meta on platforms like Facebook and Instagram. The Tournesol project aims to create a more collaborative and transparent approach to data analysis.
“We must work to develop solutions, keeping in mind that we won’t suddenly see a ‘white knight’ emerge in the tech world, but rather a multitude of solutions that are more sovereign, built in European countries, with data control and relative infrastructure control, even if we never completely disconnect from the global internet,” the source stated.
>> A propos des initiatives helvétiques : La start-up lausannoise qui défie les géants de l’intelligence artificielle, Forces et limites du nouveau modèle suisse d’intelligence artificielle Apertus, Des chercheurs suisses lancent leur modèle IA de langage ouvert and Pour développer l’industrie des puces, la Suisse finance la recherche plutôt que le secteur privé
Propos recueillis par Julie Rausis
Adaptation web: cab
