Meta's No Language Left Behind model, covering 200 languages including 55 African languages, directly eliminated funding for specialized startups across the Global South. "Investors were like, 'Facebook has solved it, so your little puny startup is not going to be able to do anything,'" AI ethics researcher Timnit Gebru told the AI Now Institute.
OpenAI representatives have approached language AI startups globally with a stark message: OpenAI will make them obsolete, according to Gebru's research. The companies offered minimal payment for data from organizations working on underserved languages.
The pattern creates market concentration in Silicon Valley while eliminating regional alternatives. When major tech firms announce broad language coverage, investors pressure smaller organizations worldwide to shut down, regardless of technical merit or efficiency advantages.
"People came along and decided that they want to build a machine god," Gebru said. "They end up stealing data, killing the environment, exploiting labor in that process."
Fellow researcher Abeba Birhane argues the 'AI for good' narrative deflects criticism. "It allows companies to say 'Look, we're doing something good! Everything about AI is not bad. And you can't criticize us,'" she said.
The critique arrives as mainstream development continues the resource-intensive trajectory. Recent launches include DeepSeek V4 from China and Nvidia's photonics investments for AI infrastructure, following the same paradigm: massive models requiring substantial computational resources.
Gebru and Birhane argue smaller, specialized models could serve specific communities and languages more effectively. Their research suggests billion-parameter systems impose environmental and labor costs that outweigh benefits, particularly for communities in Africa, Asia, and Latin America where targeted solutions might work better.
The central question is whether general-purpose models justify their resource demands and market effects. The dominant paradigm's market dynamics prevent alternatives from securing funding or partnerships globally, creating dependency on Silicon Valley infrastructure for language technology worldwide.