Investors are telling Global South AI startups to shut down after Big Tech announces multilingual models covering their languages, according to AI ethics researcher Timnit Gebru.
When Meta released its No Language Left Behind model claiming to translate 200 languages including 55 African languages, investors told small African NLP startups "Facebook has solved it, so your little puny startup is not going to do anything," Gebru said in a new AI Now Institute report.
OpenAI representatives have approached small language AI organizations with threats that OpenAI will make them obsolete, offering "peanuts" for their data, Gebru reported. The pattern repeats across Big Tech: announce broad coverage, then eliminate specialized competition.
The criticism targets the foundation model paradigm itself. "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."
Ethics researchers say the approach lacks empirical justification for its claimed benefits while creating market concentration. The "AI for good" framing serves as PR deflection from grassroots resistance movements, according to researcher Abeba Birhane.
"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,'" Birhane said in the report.
The critique comes as AI infrastructure investment accelerates. Big Tech companies are expanding data center capacity and multimodal capabilities, widening the gap between ethical concerns and commercial momentum.
Small specialized organizations often have deeper cultural and linguistic expertise for specific communities. Foundation models claim universal coverage but may lack nuanced understanding, critics argue.
The competitive dynamics create a self-fulfilling prophecy: Big Tech announces broad coverage, investors withdraw funding from specialists, then Big Tech claims to be the only viable option. Startups in the Global South face particular pressure as their focus areas get absorbed into larger models.
The researchers call for scrutiny of whether foundation models actually serve claimed beneficiaries or primarily consolidate market power. The pattern suggests strategic use of "coverage" announcements to eliminate competition before products fully deliver.