Google is once again under scrutiny after reports surfaced that Gmail’s artificial intelligence (AI)-powered translation feature is altering the content of emails in German — without users’ consent or even awareness. The issue was first identified by German journalists who discovered that email newsletters sent in German were being silently “corrected” or retranslated by Gmail’s backend AI systems, resulting in unexpected and sometimes offensive changes.
This unexpected intervention has raised serious concerns over user control, AI hallucinations, and the broader implications of machine translation embedded invisibly into essential digital services.
When AI Oversteps Its Role
Among the most striking errors reported was the automatic translation of the German word “Ass”— used in a political context to describe a strategic move — as “Arsch,” an explicit term in English. Other cases showed phrases like “Ukrainian positions” being reinterpreted as “American positions,” and references to the “Israeli army” appearing instead as “Russian army.” These are not trivial typos, but significant distortions of meaning that can have diplomatic, journalistic, and professional consequences.
The source of the issue appears to be Gmail’s difficulty in processing mixed-language content, particularly when HTML formatting includes English tags or code. In trying to make sense of these signals, the AI seems to retrigger translation functions—even when the original content is already in the user’s preferred language.
A Wake-Up Call for AI in Communication
While Google has acknowledged the issue and is working on a fix, the rollout is expected to take several weeks. For now, millions of Gmail users — especially those outside the English-speaking world — remain vulnerable to these silent alterations.
This incident highlights a broader challenge in the localization and language tech space: As AI becomes more deeply embedded in everyday tools, the potential for invisible errors grows. Human oversight, clear user settings, and transparency must remain central to how AI is deployed — especially in something as fundamental as written communication.
