Industries / Finance and business services
Translation services
Translation services put writing and speech from one language into another for businesses, courts, hospitals, governments and media, through translators, interpreters and, more and more, software whose output people check.
196 BCE
A council of Egyptian priests meeting at Memphis passed a decree honoring King Ptolemy V and ordered copies cut on stone slabs in three scripts: hieroglyphs, demotic (the everyday Egyptian script) and Greek. By the British Museum's account the Greek and hieroglyphic versions were made from the demotic. One copy, found at Rosetta in 1799 and now called the Rosetta Stone, later let scholars read hieroglyphs.
40
Milestones achieved across 6 domains.
7
To watch for: announced, not yet confirmed outside.
12
Open deliverables. Nobody has shipped them.
In one minute
Where it started
Governments have paid for translation as long as they have ruled people who speak different languages: in 196 BCE Egyptian priests issued a decree in Egyptian and Greek, and a copy survives as the Rosetta Stone. Interpreting became a modern trade when headphones let interpreters speak at the same time as the speaker, first at the International Labour Conference in 1927 and then at the Nuremberg trials. Machine translation began with a public demonstration by Georgetown University and IBM in 1954, and from the 1990s translation memory software and the translation of software and websites built a business of agencies and freelancers.
Where it stands
The analyst firm Nimdzi puts the language services market at USD 72.6 billion in 2025, growing less than 1% a year; Slator, counting more narrowly, puts it at USD 30.85 billion. Machine output is now the starting point for much of the work: at the 2025 Conference on Machine Translation a general AI model matched or beat professional translations in 10 of 16 language pairs, though human translations were back in the top group in 8 of 12 pairs in 2026, and Nimdzi reports that price pressure and the tedium of editing AI output have driven many linguists out of the profession. Interpreting, a third of the largest providers' revenue, is still done mostly by people in courts, hospitals and conference rooms, while machine interpreting reached earbuds in 2025.
What is open
Most open jobs are public tests nobody has published: how often AI tools break software text, Word layouts and subtitle rules; how lip-synced dubs, earbud interpreters and AI reviewers compare with professionals; and how well open models translate languages their makers never checked. Others are checkers built on public records published in two dozen languages, such as EU laws and EU medicine leaflets, that flag where one language version says something different from the rest. Each can be done with openly licensed material, volunteers or actors, and current AI models.
Domains
AchievedWatch forOpen
Domain
Progress
Last achieved
Watch for
Open
Document translation
Translators turn written material such as manuals, reports, websites and the text inside software from one language into another, usually working in software that offers them earlier translations to reuse.
7 · 1 · 2
Last achievedThe market stops growing2025
Watch forCanada's in-house AI translator
2
openMachine translation and editing
Computers translate text from one language into another, and people correct the output, a step called post-editing, until it is fit to use.
7 · 2 · 2
Last achievedMachines join the top group2025
Watch forOpen models built for translation
2
openInterpreting
Interpreters put speech from one language into another as it is spoken, at conferences, on the telephone and over video links, and software now tries to do the same.
7 · 1 · 2
Last achievedTranslation inside earbuds2025
Watch forLive translation in any headphones
2
openSubtitles and dubbing
Subtitlers write a film's or video's dialogue in another language as short timed lines on screen; dubbing replaces the voices with speech recorded, or now generated, in another language.
5 · 1 · 2
Last achievedAI anime dubs pulled2025
Watch forDubbed and lip-synced reels
2
openLegal and medical translation
Translating and interpreting where a mistake can decide a case or harm a patient: in courts and police stations, in the text of laws and patents, and between clinicians and patients.
7 · 1 · 2
Last achievedFederal language order revoked2025
Watch forAn in-house AI interpreter
2
openChecking quality
Checking quality is the work of judging whether a translation is accurate and fit for its purpose, by a second linguist, an examination, or a score assigned by a person or a program.
7 · 1 · 2
Last achievedLearned scores beat BLEU2022
Watch forGates that decide publication
2
open
Advances to build with
2 research notes

Claude Fable 5.2
Three working devices
Anthropic says its AI model Claude Fable 5.2 designed three prototype devices that worked without electrical modifications, with no outside test reported.

Mistral Large 4
A trillion open parameters
Mistral published its Large 4 AI model, a trillion numbers by its count, free to download and sell, though firms that rent out models need a separate license.
Open in translation services
3 of 12 open deliverables
Open · Document translation
Score AI on app strings
A public test set and scorer for AI translation of software text. Take the string files (gettext.po files) of open-source programs that already ship human translations, remove the translations, and have several AI models translate them. A script then checks every string for broken placeholders such as %s or {name}, missing plural forms (Polish, Russian and Arabic need more than English's two), damaged markup and lines over the length limit, and native-speaking translators grade the meaning of a sample without knowing which model wrote it.
Built with what exists
.po files downloaded from GNOME's translation site (l10n.gnome.org) and other open-source projects; the Translate Toolkit's pofilter, which already checks printf placeholders, variables, XML tags, plural counts and length; AI models such as Claude Fable 5.2 (see the Note "Three working devices") and the open-weight Mistral Large 4, which can run on a team's own servers (see the Note "A trillion open parameters"); Python.
Open · Machine translation and editing
Rerun the editing-time study
A repeat of the 2022 DivEMT study with today's AI translation systems. Professional translators post-edit machine translations of the same English documents into Arabic, Dutch, Italian, Turkish, Ukrainian and Vietnamese, and translate some documents from scratch as a baseline, while an editor program logs every keystroke, pause and second. Publish the logs and the seconds per word for each system and language beside the 2022 figures.
Built with what exists
DivEMT's released documents, logs and analysis code (github.com/gsarti/divemt); GroTE, the open logging editor built for QE4PE; output from current systems such as Gemini, DeepL, Google's open TranslateGemma and Mistral Large 4, whose published weights let anyone rerun the same version later (see the Note "A trillion open parameters").
Open · Interpreting
A repeated interpreting scorecard
A public scorecard that plays the same recorded European Parliament speeches through the live translation tools people can buy, including the earbud and phone features above, and compares their output with the official human interpretation of the same speeches, scored for meaning errors and delay, and rerun every six months so the trend shows.
Built with what exists
The ESIC corpus of Dominik Macháček and colleagues (2021), 10 hours of English European Parliament speeches with their interpretation into Czech and German; European Parliament plenary recordings with their interpretation channels; Apple Live Translation and Google Translate's live beta; Pan and Seeber's reference-free scoring, calibrated against expert ratings; a speech-to-text model such as ElevenLabs Scribe v3 to transcribe both the machine and the human output (see the Note "Fewer word errors"); two professional conference interpreters per language to grade blind.