Every day, billions of people use AI-powered translation, unaware of its inherent shortcomings, to communicate, access services, and make decisions. But when these translations fail, the consequences can be serious—affecting health, justice, safety, and fundamental rights. Despite this, most global AI safety frameworks remain silent on translation, treating it as a solved or low-stakes problem. It isn’t.
Our new report, Why Translation Errors Must Be Part of the AI Safety Discussion, shows how hidden risks—especially correspondence errors, where distorted meaning is hidden under readable text—can lead to real-world harm. From medical instructions that put patients at risk to asylum claims lost in mistranslation, the evidence is clear: translation errors are a high-impact AI safety issue. The report also presents a practical solution: a label-first model that makes the difference between Professionally Verified Translation and Unverified Translation visible to everyone.
Translation touches every corner of society, and we cannot afford to treat it as an afterthought in AI governance. Whether you’re a policymaker, technologist, service provider, or advocate for equity, this report lays out the path to safer, more accountable cross-language communication.
To download a copy of the report, please click on the image of the report’s cover.