Investors told small language AI startups to "close up shop" after Meta announced its No Language Left Behind model covering 200 languages, including 55 African languages, according to AI ethics researcher Timnit Gebru.
"[Investors] were like, 'Facebook has solved it, so your little puny startup is not going to be able to do anything,'" Gebru said in a report published by AI Now Institute.
OpenAI took a more aggressive approach with organizations working on non-English languages. "OpenAI is going to put you out of business soon because we're going to make our models better in your language. You're better off collaborating with us and supplying us data for which we're going to pay you peanuts," representatives allegedly told small language organizations, Gebru reported.
The pattern repeats globally when any major tech company announces a large multilingual model. Investors immediately pressure smaller organizations working on specific languages to shut down operations.
Gebru and fellow researcher Abeba Birhane are challenging what they call the "one model for everything" paradigm pursued by Silicon Valley tech giants. "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."
The researchers critique the "AI for good" framing used by major tech companies to deflect criticism. "[AI for good] is a way to paint a positive image of AI technologies, especially in light of a lot of the backlash—like the resist or refuse AI grassroots movement that's emerging," Birhane explained. "'AI for good' allows companies to say 'Look, we're doing something good! Everything about AI is not bad. And you can't criticize us.'"
The dynamic threatens linguistic diversity in AI systems worldwide. When investors abandon small language startups, communities lose organizations with deep cultural and linguistic expertise in favor of one-size-fits-all models that may fabricate outputs or misrepresent languages.
The researchers advocate for "frugal AI"—models built for specific tasks and languages rather than general-purpose systems. This approach could preserve space for small language organizations that understand local contexts better than global tech companies.