Language is the missing link in the Al revolution

These were the thoughts running through my mind on the train to London AI Week where I am slated to speak about a future where AI truly serves humanity. Around me, people spoke in German, Polish, Tamil, and English. The mix of languages reminded me that while the world grows more connected, the digital world remains linguistically lopsided. If you asked your phone in Polish for a definition of "multimodal AI," you might get a passable response. Ask in Tamil, and it may struggle. Ask in Quechua or Wolof or Karay-A, and you'll likely get silence.

- NIGHT OWL ANNA MAE YU LAMENTILLO

This isn't just a technical quirk—it's a quiet crisis. We've made tremendous progress expanding internet access, improving device affordability, and building digital skills across the globe. But one enormous gap remains: most of the world cannot speak to AI—and cannot be heard by it—in their own language.

There are 7,159 living languages spoken today. As of 2023, fewer than one percent of them are fully supported by AI systems. Only 32 languages were classified as “digitally thriving” in Ethnologue’s Global Digital Language Support Scale. Just 108 more were deemed “vital.” That leaves over 7,000 languages with limited or no access to speech recognition, machine translation, text-to-speech tools, or even reliable typing interfaces.

This means that while AI is marketed as a universal tool, it is still speaking mostly to the privileged linguistic elite. The majority of humanity remains on mute.