# AI translation improves but human checks still needed for high-stakes text, experts say

Digest AI · Generative AI & Models · published 2026-10-13T23:05:29Z

Canonical: https://digestai.news/story/ai-translation-improves-but-human-checks-still-needed-for-high-stakes

## Summary

AI translation now handles over one trillion words a month on Google Translate alone, with the standalone machine-translation market worth an estimated $1.4 billion last year. For low-risk tasks such as internal documents or product listings, raw AI output is often sufficient, says John Zimmerman of the Swiss non-profit OKA, which uses AI drafts plus human review to publish thousands of Wikipedia articles. However, specialists warn that errors remain dangerous in legal, medical and asylum contexts. Translator Ariella Baltié found AI mistranslating "Le demandeur" as "the requester" and "épouse" as "marries" in court documents. Medical translator Carol Bereuter cited a case where the Greek letter μ was read as milli instead of micro, creating a thousand-fold dosage error. Researchers note that fluent-sounding output makes mistakes harder to spot. An OKA study of 119 articles across 10 language pairs found 27% of AI text still required correction, with 5-6% of errors changing the original meaning. CNRS research director François Yvon says current probabilistic models make 100% reliability unlikely, and the real breakthrough will come when systems can signal uncertainty.

## Key points

- Google Translate processes over one trillion words monthly; standalone machine translation market worth $1.4bn
- OKA study: 27% of AI drafts corrected, 5-6% of errors changed meaning across 119 articles and 10 language pairs
- Experts cite legal, medical and asylum errors; fluent output makes mistakes harder to detect

## Why it matters

AI translation is cheap and fast for routine use, but high-stakes fields still require human verification because fluent errors can have serious legal, medical or humanitarian consequences.

## Sources

1. [How long will humans be needed for translation?](https://bbc.co.uk/news/articles/c1wxdjx27ypo?at_medium=RSS&at_campaign=rss) (BBC Technology, 2026-10-13)

## Cite

Digest AI, "AI translation improves but human checks still needed for high-stakes text, experts say", 13 October 2026, https://digestai.news/story/ai-translation-improves-but-human-checks-still-needed-for-high-stakes

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/ai-translation-improves-but-human-checks-still-needed-for-high-stakes.json