AI’s Dirty Secret: It Mostly Speaks English | Forbes Technology Council Article
AI is often described as a global technology, but most leading AI models are still disproportionately built, trained, and evaluated in English. This creates a significant challenge for models builders and organizations developing AI systems for international audiences.
In her latest Forbes Technology Council article, DATAmundi CEO Véronique Özkaya explores why language in AI is about much more than translation. AI systems must also understand cultural context, tone, politeness norms, and the subtle ways people communicate in different languages and markets.
Read the full Forbes Technology Council article by Véronique Özkaya: “AI’s Dirty Secret: It Mostly Speaks English.”
Article Highlights
- Modern AI has been built on English. Studies such as Meta’s LLaMA 2 paper indicate that roughly 90% of training tokens are English, while broader web data suggests English still accounts for nearly half of online content.
- Multilingual doesn’t always mean culturally aware. Models can generate technically correct responses while missing important cultural nuances.
- Bias toward English limits global adoption. Many of the fastest-growing online populations are in regions where English is not the primary language.
- High-stakes applications require local understanding. In domain-specific areas such as healthcare, finance and legal services, misunderstanding local communication can have serious consequences.
- The human data layer remains essential. Native-language content, linguistic expertise and culturally grounded annotation are critical for building and training AI that performs reliably across languages and regions.
- The next frontier isn’t more data, it’s better data. Working with representative datasets are key to creating truly global AI systems.
Building AI for the world requires more than translating English-first systems. It requires designing for linguistic and cultural diversity from the outset.
Read the full Forbes Technology Council article by Véronique Özkaya: “AI’s Dirty Secret: It Mostly Speaks English.”
