Translation noise is the small distortion, omission, or word choice error that can appear when text is machine-translated. Even fluent output may shift emphasis, simplify ambiguity, or mistranslate technical terms. In cybersecurity, that matters because security work depends on precision: a single changed word in an alert, policy, indicator, or incident note can alter how people respond.
Attackers can exploit translation noise by hiding instructions in multilingual phishing, scams, or support messages, hoping the translated version will soften warning signs or change intent. Defenders also have to manage it in multilingual SOCs, where translated tickets and threat reports may be reviewed as if they were exact copies of the source. Good practice is to keep source text with the translation, record model or tool version, and add human review for legal, incident-response, and customer-facing content.