Meta's No Language Left Behind model covering 200 languages, including 55 African languages, triggered immediate funding cuts for African NLP startups. Investors told founders to close operations because Facebook had 'solved' the problem, according to AI ethics researcher Timnit Gebru.
OpenAI representatives allegedly threatened small language AI organizations across the Global South, claiming the company would make them obsolete. 'OpenAI is going to put you out of business soon because we're going to make our models better in your language,' researchers report the company telling startups while offering minimal data compensation.
This pattern repeats globally wherever Big Tech announces broad multilingual capabilities. Local organizations with language-specific expertise and cultural knowledge lose funding despite questions about whether massive general-purpose models actually perform as claimed.
Gebru challenges the scaling paradigm's premise: 'People came along and decided that they want to build a machine god and then claimed that they are doing it. They end up stealing data, killing the environment, exploiting labor in that process.'
Co-researcher Abeba Birhane criticizes 'AI for good' framing as deflection. 'It allows companies to say: Look, we're doing something good. Everything about AI is not bad. And you can't criticize us,' she states.
Safety evidence remains thin even in critical applications. Medical transcription systems using large language models produce hallucinations affecting patient care. Evidence for claimed AI benefits in healthcare and other domains lacks empirical support.
The analysis, published by the AI Now Institute, questions whether resource-intensive general-purpose models serve actual needs or primarily consolidate Big Tech's market dominance at the expense of local innovation in developing regions.