On a cold February stage in Paris, Google CEO Sundar Pichai did something rare for a tech executive: he acknowledged the quiet success of an international diplomatic campaign. “We added over 110 new languages to Google Translate last year,” he said, citing progress that seemed technical but was, in fact, deeply political.
This wasn’t just about code. It was about culture. And for Joseph Nkalwo Ngoula, a digital policy advisor at La Francophonie’s UN mission in New York, it marked a subtle turning point: big tech was finally listening.
For two years, La Francophonie — an organization representing 93 states and governments that use French — had been pushing for linguistic diversity in AI. Not just more languages, but better representation of non-English speakers in the tools shaping our digital future.
Why It Matters
While only 20% of the world speaks English at home, nearly half of the training data for AI models like GPT-4 and Gemini is in English. This isn’t just a mismatch—it’s a digital power imbalance. Non-English queries still yield weaker answers, and hybrid or regional dialects? Forget it. Ask ChatGPT a question in Camfranglais, and you’re likely to get a shrug.
La Francophonie’s campaign turned this frustration into policy. Behind the scenes, they leveraged diplomatic alliances—even gaining support from Lusophone and Hispanic blocs, and unexpectedly, from the U.S.—to influence the UN’s Global Digital Compact. When it was adopted in late 2024, the Global Digital Compact included a clear call for cultural and linguistic inclusion in AI development.
Still Just Words?
Not quite. Pichai’s speech in Paris and Google’s updated Translate platform show some traction. But advocates warn that structural issues persist. Algorithms still prioritize English-language content; regional variants are often flattened. Even languages like French are treated as static, textbook constructs.
“Molière, Senghor, Césaire—they’d all be turning in their graves if they saw how A.I. writes French today,” Ngoula quips.
The Stakes
The stakes are cultural survival in the digital age. Without robust datasets in diverse languages, AI systems will continue to hallucinate, fabricate, or ignore non-dominant narratives. The danger isn’t just bad translations—it’s epistemic erasure.
For Ngoula and his peers, the goal now is clear: keep pushing. Linguistic diversity must not be a checkbox—it has to be the backbone of digital policy. Because in the world of AI, language isn’t just how we communicate. It’s how we exist.
And if the algorithms are going to represent us, they’d better learn to do it sharp-sharp.
