Tag: machine translation

The Commercial Promise Versus Courtroom Reality

Commercial language service providers frequently present barristers and solicitors with an appealing narrative regarding the maturity of translation technology in international litigation and transactional work. A representative example appears in a 2023 contribution to Barrister Magazine, which contends that machine translation saves considerable time and money across high-volume caseloads, that output accuracy depends largely on database size and customised engine training, and that pairing raw machine output with human post-editing guarantees 100 per cent accuracy for instructing counsel. For legal practitioners under pressure to review foreign disclosure bundles or cross-border transactional documentation within compressed timetables, this hybrid workflow is marketed as a reliable compromise between automated speed and human assurance.

A technorealistic assessment of legal translation technology requires separating low-risk information management from high-stakes legal representation. Neural machine translation engines and generative artificial intelligence models can assist legal teams with initial document triage or internal gisting when counsel simply need to ascertain the general subject matter of a foreign file. Relying on machine output as the foundational draft for evidential documents, statutory provisions, or binding commercial agreements introduces systemic risks that post-editing struggles to remediate. Empirical research in legal translation studies and comparative law, particularly across the Greek civil law and English common law pair, demonstrates that each pillar of the commercial post-editing narrative rests on a flawed conception of how legal language functions across jurisdictions.

Why Database Size and Customised Engine Training Fail to Guarantee Legal Accuracy

The commercial argument for hybrid legal translation rests heavily on the assumption that translation quality is a function of corpus volume and continuous machine training. Proponents suggest that feeding millions of bilingual legal texts into a neural engine, combined with ongoing corrections from human editors, equips the software with a substantive grasp of legal terminology and phrasing. This premise overlooks the fundamental architecture of both neural machine translation and large language models. As Argyri Panezi and I (2026) observe alongside legal technology scholars such as Eliza Mik (2026), automated systems process tokens on the basis of statistical probability and distributional frequency across training corpora. They lack any capacity to comprehend legal meaning, legislative intent, or jurisdictional consequences.

Legal terminology is inherently system-bound, as Susan Šarčević (1997) and Deborah Cao (2007) have long established in comparative legal linguistics. Every national legal order develops its own conceptual architecture shaped by distinct historical and doctrinal evolutions. In their corpus-driven investigation of Greek notarial deeds and their English translations, Vilelmini Sosoni and John O’Shea (2021) examined the transfer of property law concepts between Greek civil law, which draws on Roman-Byzantine traditions and the German Civil Code (Bürgerliches Gesetzbuch), and English common law, which remains rooted in feudal principles of Crown landholding and estates in land. Across the ten most frequent property law terms extracted from twenty Greek notarial conveyancing deeds, Sosoni and O’Shea found zero instances of exact or near equivalence between Greek and English. Half of the core terms exhibited only partial equivalence, while the remaining half represented complete conceptual non-equivalence, requiring careful descriptive paraphrases or functional adaptations grounded in comparative legal analysis.

Empirical testing confirms that expanding a legal database fails to resolve this inter-systemic incongruity. In a recent comparative evaluation, Vilelmini Sosoni and John O’Shea (2026) tested a domain-adapted commercial neural machine translation engine alongside leading generative artificial intelligence models (ChatGPT-4o, Gemini 1.5 Flash, and Claude 3 Haiku) on two authentic Greek legal texts: an extract from Greek planning and environmental legislation (Law 3937/2011 and Law 4280/2014) and a notarial property sale agreement. The domain-adapted neural engine had been trained on a substantial parallel corpus of 10.5 million sentence pairs, of which 5.3 million sentence pairs (50.5 per cent) consisted specifically of Greek-to-English legal and legislative texts.

Despite this extensive domain-specific training, the customised legal engine recorded the worst overall severity index of all systems evaluated across both the legislative text (4.54) and the property sale agreement (4.33). It also produced the highest proportion of Critical accuracy errors, reaching 21.05 per cent in the statutory extract and 10.26 per cent in the conveyancing contract. The generative artificial intelligence models similarly struggled with legal precision, with accuracy errors accounting for between 51.11 per cent and 60.38 per cent of all annotated errors across every model in the legislative text.

The qualitative failures documented in Sosoni and O’Shea (2026) illustrate the specific legal liabilities created when statistical token prediction encounters system-bound concepts:

  • Polarity reversals in statutory provisions: In the legislative text, the domain-trained legal engine inverted the operative force of a mandatory rule by translating the Greek verb ισχύει (“applies to”) as “shall not apply to”. In litigation or regulatory compliance, reversing a statutory prohibition into an exemption alters the entire legal position of a party.

  • Temporal shifts in contractual warranties: In the Greek property sale agreement, the vendor warranted that the real estate was presently free of encumbrances and third-party claims (ελεύθερα). The domain-adapted engine rendered this present warranty of title as a future obligation (“shall be free and exempt”), altering the allocation of transactional risk at completion.

  • Conflation of civil and criminal legal institutions: Every neural and generative system evaluated in the study mistranslated the Greek civil enforcement term κατάσχεση (civil attachment or distress levied against property for debt recovery) as “seizure”, importing a criminal or public-law concept into a private conveyancing instrument.

  • Conceptual flattening of property rights: Distinct Greek civil law concepts were routinely reduced to vague everyday expressions. The term κατοχή (physical detentorship under the Greek Civil Code, as distinct from legal possession or νομή) was flattened into “occupancy” or “control”, while οροφοκτησία (horizontal ownership of an entire floor under Law 3741/1929) was stripped of its statutory specificity and rendered simply as “ownership”.

These findings converge with empirical research by Stavroula Paraskevi Vraila (2026), who analysed 400 translations of system-bound legal terms into Greek across DeepL, Google Translate, GPT-5.3, and Claude Opus 4.5. Vraila demonstrated that because commercial models rely overwhelmingly on Anglophone training corpora or English pivot representations, they exhibit high rates of English-mediated terminological displacement, averaging 57.9 per cent across neural engines and 28.5 per cent across generative models. Fundamental Romanistic property distinctions between νομή (possession) and κυριότητα (ownership) were systematically collapsed by the majority of engines into imprecise common law approximations, substituting the doctrinal categories of one legal tradition for those of another.

The Post-Editing Myth: Deceptive Fluency and the Absence of Productivity Gains

When commercial providers concede that raw machine translation cannot be trusted in isolation, they invariably present human post-editing as the safeguard that restores 100 per cent accuracy while preserving time and cost savings. This claim ignores a well-documented cognitive hazard in neural translation workflows known as deceptive fluency. Earlier statistical machine translation produced visibly broken syntax that immediately alerted a reviewing lawyer or linguist to underlying mistranslations. Contemporary neural engines and large language models are optimised through alignment techniques such as reinforcement learning from human feedback to generate smooth, idiomatic target prose (Ouyang et al., 2022; Vraila, 2026). As Elisabeth Wiesmann (2019) and Karolina Stefaniak (2020) have shown, grammatical fluency conceals severe semantic departures from the source text, inducing automation bias in human reviewers who become far less likely to detect subtle omissions, fabricated qualifiers, or shifted legal obligations.

Empirical process research refutes the claim that post-editing delivers efficiency or superior accuracy in specialised legal translation. In an experimental study involving professional translators working from Greek into English on property law texts, Vilelmini Sosoni, John O’Shea, and Maria Stasimioti (2022) measured temporal, technical, and cognitive effort using keylogging and eye-tracking data to compare translation from scratch against the post-editing of neural machine translation. The investigation revealed zero productivity gains from post-editing in legal texts. Furthermore, working over raw machine output introduced substantial cognitive friction: translators primed by the machine draft committed more verity and style errors than when they constructed the legal translation directly from the Greek source text.

The quantitative metrics from Sosoni and O’Shea (2026) explain why repairing machine output consumes so much professional effort. Across the four neural and generative systems tested on the Greek legislation and the property sale agreement, Translation Edit Rate (TER) scores ranged between 56.56 per cent and 74.42 per cent. A human reviewer confronted with a machine draft in which well over half to nearly three-quarters of the text requires structural modification must spend extensive time unpicking plausible-sounding errors before reassembling the syntax to match English legal drafting conventions.

In the contractual text analysed by Sosoni and O’Shea (2026), style errors alone accounted for between 41.38 per cent and 43.75 per cent of all errors produced by the generative artificial intelligence models. Vraila (2026) similarly found that stylistic flattening affected between 18 per cent and 85.7 per cent of legal outputs across systems, eroding the formal register conventions that carry performative force in legal instruments. Unravelling a flattened, conceptually displaced machine draft demands prolonged verification against the source text, eliminating the promised commercial speed while increasing the probability that a camouflaged terminological error survives into the final court bundle or closing set.

The Human-in-the-Loop Fallacy, Professional Liability, and Client Confidentiality

The commercial promotion of hybrid workflows relies heavily on the reassuring label of the “human-in-the-loop”. In their examination of legal liability and ethical governance in automated translation, Argyri Panezi and John O’Shea (2023) demonstrate that this phrasing misrepresents the operational and legal reality faced by the specialist translator. When a lawyer-linguist translates a pleading or contract from scratch, they exercise full epistemic and professional control over every interpretive decision from the opening recital to the execution clause. When placed at the end of an automated pipeline to post-edit output generated by an opaque neural architecture, the linguist forfeits control over the cognitive baseline of the draft while remaining expected to assume full contractual and professional liability for the accuracy of the delivered text.

Practising lawyers who handle bilingual legal documentation recognise these hazards in their daily work. In a survey-based empirical study examining how translating lawyers in Greece adopt and perceive machine translation, Kyriaki Kourouni and John O’Shea (2025) found that two-thirds of respondents avoided machine translation altogether, electing to translate legal texts manually. Among the minority of legal practitioners who did use machine translation, 61.55 per cent restricted its use to occasional, low-stakes tasks, such as obtaining a rough gist of a foreign text (9.28 per cent) or drafting routine correspondence with international clients (8.25 per cent). Practitioners rejected the notion that automated output could serve as a reliable basis for court submissions or binding transactional instruments.

The survey by Kourouni and O’Shea (2025) also highlighted severe vulnerabilities surrounding data protection and professional ethics. Although commercial articles frequently caution against free web-based translators, 73.33 per cent of the surveyed lawyers who used machine translation relied on free Google Translate. While 35.72 per cent of respondents acknowledged that uploading client texts to external servers risked breaching confidentiality obligations, far fewer connected the use of automated translation to professional negligence, malpractice exposure, or breaches of the duty of care owed to clients (Wahler, 2018; Giampieri, 2023). Even where paid application programming interfaces offer contractual assurances regarding data retention, feeding sensitive corporate disclosures or litigation strategies into third-party generative models raises unresolved questions under data protection regimes and professional conduct rules.

Systemic Language Pollution and the Efficiency Paradox in Cross-Border Practice

The widespread marketing of machine translation as a routine component of legal case management carries systemic consequences that extend beyond individual translation errors. In their analysis of artificial intelligence and legal epistemology, Argyri Panezi and John O’Shea (2026) introduce the concept of “language pollution” to describe how unverified machine translations, fabricated citations, and algorithmically mediated legal transplants degrade the linguistic and doctrinal integrity of the legal system. Because law is constituted through authoritative texts and precise interpretive traditions, injecting statistically generated approximations into legal workflows contaminates the shared informational environment relied upon by counsel, transactional parties, and the judiciary.

Rather than streamlining cross-border representation, uncritical reliance on automated language tools triggers an efficiency paradox in legal practice (Evans, 2025; Panezi and O’Shea, 2026). When clients, overseas instructing solicitors, or commercial agencies use generative models or hybrid machine workflows to translate and interpret foreign law, they frequently construct entirely false legal premises that specialist counsel must subsequently dismantle.

Panezi and O’Shea (2026) document a telling empirical demonstration of this phenomenon. When asked in June 2025 and again in November 2025 to identify the Greek legal rules governing the construction of boundary fences between neighbouring properties in Athens, ChatGPT produced confident, structured legal memoranda citing Articles 1003 to 1010 of the Greek Civil Code. Upon inspection against the actual statutory text, those provisions have no bearing on the erection of boundary walls: Articles 1003 to 1005 regulate emissions of smoke, soot, and noise; Article 1006 addresses the risk of collapsing structures; Article 1007 governs digging near a boundary; Article 1008 deals with encroaching tree roots and branches; Article 1009 regulates fruit falling onto a neighbour’s land; and Article 1010 addresses construction encroaching onto neighbouring land in good faith. When pressed for supporting authority, the model fabricated non-existent case law from the Greek Supreme Civil and Criminal Court (Areios Pagos) alongside invented planning statutes. When machine-generated translations or summaries of this kind enter a cross-border file, experienced practitioners lose substantial time correcting fabricated statutory frameworks and re-educating clients before genuine legal work can proceed.

The Lawyer-Linguist Standard in Cross-Border Legal Practice

The empirical record from Greek-to-English legal translation research refutes the commercial narrative that larger databases, customised engines, and human post-editing can replicate the work of a specialist legal translator. Statistical models cannot bridge the conceptual divide between civil law and common law systems because resolving terminological incongruity requires comparative legal reasoning rather than probabilistic token matching. Post-editing fluent machine prose fails to save time in complex legal genres and exposes legal teams to camouflaged polarity reversals, displaced property and procedural concepts, and unquantified professional liability.

Cross-border litigation, corporate acquisitions, and real estate conveyancing involving Greece and Cyprus require a translation partner grounded in the substantive law of both jurisdictions. At JurTrans, every instruction is handled directly by a specialist lawyer-linguist with deep familiarity with the nitty-gritty of the legal systems involved, thinking and writing like a lawyer to ensure that complex civil law concepts are rendered into precise, authoritative English. Through rigorous attention to detail and direct human scholarship from the first word to the final review, JurTrans delivers translations that are entirely free of fabricated content and fit for the demands of the Bar and international legal practice.

Editorial note: This article examines empirical research on translation methodology and technology risks for general informational purposes; nothing written herein constitutes legal advice.

Does reputation matter?

What happens when a supreme court plugs MT into its website?

Check out the images below:

errors in translated name of supreme court errors in translated name of supreme court errors in translated name of supreme court

Remember we are talking about Greece’s most important court. Not just any old court.

Note that the court’s full and proper name already appears on its website in English.

It’s the MT plug-in that consistently gets it wrong.

The versions that appear in the reel are just some of the many different variants that it comes up with.

Words in law matter – even translated words

Critics will say, he’s going on again about MT, and that the translation is provided for information purposes only so there’s nothing to worry about.

Critics will also say that the court is offering the public a service, doing us a favour.

Are they?

Nothing could be further from the truth.

In fact I’d say it’s a disservice. To reiterate: Words in law matter – even translated words

If the name of the court (something simple and straightforward) can’t be got consistently right, what does that say about the rest of the content generated?

Can anyone reading the machine-generated output think this is useful or helpful to them in any way?

When is a “service” not a service?

Let’s delve a little deeper into the idea that this is a service.

If you boil things down, a service is something you seek out, you pay for (typically) and which adds value.

A service involves doing something for someone that is valued.

It involves applying skill, competence and expertise for the benefit of another.

Providing automatically-translated versions of court judgments, while well-intentioned, hardly meets those requirements.

Broader considerations

It also raises several important legal policy and access to justice considerations.

Even if such translations are labelled as “for information purposes only” (often disclaimers like that are missing) and are viewed by some as better than no translation at all, there are valid counterarguments to consider:

Legal and Ethical Responsibility: Courts have a responsibility to ensure that their communications are clear, accurate, and accessible in the language of the court.

Why should that be any different if the court opts to provide translations?

Offering substandard translations could be seen as neglecting this responsibility, potentially undermining public trust in the judicial system.

Reliance and Legal Consequences:  Relying on machine translations for legal decision-making, even when they are marked as “for information purposes only,” poses significant risks, especially for those without access to professional translation services.

Court judgments, like other legal documents, are filled with complex terminology and nuanced language that machine translation often fails to accurately capture.

This can lead to misinterpretations about legal rights, obligations, and the judgment’s implications, resulting in incorrect decisions or unnecessary time and expense spent consulting legal advisors to correct misunderstandings.

Equal Access to Justice: Access to justice implies that everyone, regardless of language proficiency, should have equal access to legal information.

By providing low-quality translations, non-native speakers are disadvantaged, potentially violating the principle of equality before the law.

Recommendations

To address these issues, several recommendations could be considered:

Clear Disclaimers and Guidance: Provide clear disclaimers about the limitations of machine translations and guiding individuals to seek professional translation or legal advice for critical matters. 

Improving Translation Quality: Invest in higher-quality translation services, potentially combining machine translation with human review and editing, to ensure accuracy.

Consult with experts in legal translation: Develop a policy for your legal translations that helps reduce your exposure to reputational risk

I’m very happy to be able to share a post about machine translation, its use in professional settings and related questions of  liability, written by Wojciech Woloszyk,  a Polish lawyer linguist. Much obliged to him for drawing my attention to it and for allowing me to share it.

It concerns a recent case before the Polish courts that looked at liability for the improper use of machine translation. While the case did not involve legal translation (it involved translation of a book), it raises interesting points about how translation services are provided.

The original post in English can be accessed here but is also reproduced in full below with Wojciech’s kind permission.

LSP’s responsibility for the process of translation and the rules of using machine translation

THE PRECEDENT-SETTING RULING OF THE POLISH COURT IN A CASE ON THE USE OF MACHINE TRANSLATION IN THE PROCESS OF PROFESSIONAL TRANSLATION AND QUALITY SUPERVISION OVER SUCH PROCESS EXERCISED BY A TRANSLATION COMPANY

The use of ‘free’ applications and services provided by global technology giants, known collectively as GAFAM or BIG TECH, for the purpose of professional activities invariably spark many controversies, disputes, and doubts. On the one hand, we all understand that, in fact, there are no free services or goods. When we are offered any type of ‘free’ and ‘gratuitous’ products, a question should be asked: ‘With what have I paid for this, if not with money? If somebody else has paid, why is that and what will they want from me?’.

It should be noted that these observations concern business entities, as we may encounter real disinterestedness and gratuitousness, for instance, in the case of philanthropy or charity.

If, however, we do not belong to a group of people who are actually in need, but we use services offered by third parties for gainful employment or business activity purposes, we should seriously ask ourselves where the catch is. Because there always is a catch… Unfortunately, for some mysterious reasons, many professionals are still in denial when it comes to the acknowledgement of this simple truth. The same happens to translators using the Google Translate service for the purpose of providing specialised translation service.

Usually, the currency we pay for ‘free-of-charge’ solutions and applications is our privacy and data on our shopping behaviours and preferences. If we accept this deal in our private life, this is basically our own choice. However, if we use such solutions and applications as part of our professional activity, these are data, content, and intellectual property rights owned by our clients and third parties that become the currency. We are not entitled to make such choices on their behalf and this type of conduct should be regarded as falling into the category of ‘unlawful act’ and ‘breach of contract’.

This issue has proved to become one of key aspects examined by the District Court for Poznań-Stare Miasto in Poznań, 12th Economic Division, in Case No. XII GC 669/17. The proceedings led to the delivery, on 13 August 2020, of a precedent-setting ruling (with the statement of reasons available HERE), which – according to my best knowledge supported by research done in commercial legal information databases and the common court case-law database held by the Polish Ministry of Justice – is the first Polish judgment referring to the use of free-of-charge machine translation tools in the context of confidentiality obligation and intellectual property rights, proper organisation of translation process, the liability of a translation company for quality supervision of a translator’s work, as well as lack of grounds to rely on a low price to justify the fact of falling short of quality requirements. The ruling also attempted to define the notion of a ‘professional translator’.

This article will discuss the most important theses included in the statement of reasons. As a preliminary remark, however, it is worth recapitulating the factual circumstances of the case to take a bigger picture.

The proceedings were initiated by a translation company (hereinafter referred to as the ‘TC’ or the ‘claimant’) which brought legal action against its client to seek the payment of remuneration for the translation of a book from Polish into English. The translation service at issue was provided in 2013. The client refused to pay and challenged the quality of the translation of its book, being a specialised handbook on the C programming of microcontrollers, provided by the TC. The translation was also very delayed. In spite of a complaint submitted and several attempts to correct the translation, the quality of the final output, according to the client, was so low (‘tragic’) that the client decided to withdraw from the agreement and ordered the translation from another entity. Against this background, a dispute arose regarding the payment of remuneration to the TC.

It should be noted that the client of the TC had a say in choosing a translator to translate the book as several translation samples prepared by different translators were presented to the client and the client chose the translation that satisfied its expectations to the largest extent. The TC assigned the translation of the book to the author of this version. The translator in question was a student of the 5th year of a Computer Science university programme and a laureate of the national competition in English language for high school students.

During the hearing, the translator stated that he had done 92% of the translation with the use of the Google Translate tool and had corrected this translation afterwards. The remaining 8% were translated with the translator tool without any verification of the output. One of the reasons for using the Google Translate tool was a delay in translating the text.

The TC and the client entered into an agreement on the translation in so-called STANDARD option, which included a reservation that the translation did not need to comply with PN-EN 15038:2006 standard (a predecessor of ISO 17100:2015), but the service would be performed by a professional translator and edited by a native speaker possessing specialised knowledge in the field of the translated text. The translation output was supposed to be fit for professional use.

An expert witness has been called. The expert witness stated that the service had been performed inconsistently with the agreement entered into and that the rules of art had been violated. The translation assessed by the expert witness included numerous linguistic errors, with some parts left untranslated, while the register of the text was far from the register expected in the case of any text intended for publication.

Having heard the case, the District Court for Poznań-Stare Miasto delivered a ruling in which it dismissed the claim and ordered the claimant to pay the cost of proceedings. The section below presents the main findings and arguments presented by the Court in the statement of reasons as well as my commentary.

Principal theses of the ruling:

  • The Court decided that the translation provided by the TC, being the claimant party, was defective to such an extent that it could not be deemed ‘fit for professional use’, contrary to what had been promised by the scope of the ‘standard’ service.
  • The translation was entrusted to a person without relevant qualifications, professional skills and experience.
  • Editing work was incomplete and negligent.
  • Additionally, the Google Translate tool was used, which constitutes a gross violation of the rules of art and contractual provisions, while the fact that the use of the tool was allowed is, at least, a flagrant example illustrating the lack of supervision of the translation process and lack of applying any quality control procedures. The use of the Google Translate tool leads to violating confidentiality obligation and intellectual property rights to the source text owned by the client or third parties.
  • The fact of leaving, in the edited text, some untranslated words, obvious linguistic and spelling errors, as well as the presence of excerpts that are entirely incomprehensible, is equivalent to the occurrence of critical errors that make it impossible to use the translation for professional purposes.
  • ‘Comprehensibility’ of the text is a minimum criterion in the case of ‘budget’ translations offered for informative (indicative) and client’s internal purposes. In the case of translations that are supposed to be used for professional purposes, the very fact that the text is ‘comprehensible’, while at the same time being unreadable, incomplete and abundant in linguistic defects, do not justify considering the translation as performed properly. The translation output in the ‘standard’ option, even after taking into account all reservations made by the TC (the claimant), should be still fit for professional use, and this quality criterion has not been met by the text in question.
  • The TC cannot rely on the contractual limitation of liability for improper performance of the agreement as the high degree of defectiveness makes the translation unfit for its intended purpose, and this, in turn, means that what we face here is not improper performance, but failure to perform the agreement (non-performance of the agreement).
  • The Court also noted that it was the claimant party itself that had quoted the price for the translation. Therefore, the claim that the price for the service justified the lower quality of the translation is groundless. The TC voluntarily agreed to provide a ‘standard’ translation service and set the remuneration itself, and therefore, it should perform the agreement in accordance with its provisions.
  • The Court found that the work provided by the TC, being the claimant party, had undoubtedly included material defects, i.e. defects that prevented its intended use or constituted an express breach of the agreement. The translation provided by the TC was useless for the defendant. This uselessness is further confirmed by the fact that the defendant was forced to re-order the translation from a third party, which did the work from scratch.
  • As a result of establishing that the work provided by the TC included material defects, the TC’s claim for payment was considered unenforceable on the ground that the agreement had not been performed, and therefore, the claim was dismissed.

Contractual limitation of the translation company’s liability for improper performance of a translation service

An extremely interesting aspect of the case was the Court’s examination of whether the contractual limitation of the TC’s liability is effective in the context of the material defectiveness of the work. The Court emphasised that, pursuant to their terms of service, the TC bears no liability for improper performance of the translation service if the client orders the translation for publication or distribution purposes within ‘standard’ and ‘econo’ options. At the same time, the Court held that the limitation of liability covered exclusively the cases of improper performance of the agreement. Given the gravity of the breach of obligations on the part of the claimant in this case, its conduct should be seen not as an improper performance but rather as non-performance of the agreement. The work provided by the claimant included material defects and, therefore, was not delivered by the TC in line with its obligations. The client refused to accept the work from the TC, which it was entitled to do. Therefore, the TC did not perform the agreement improperly but failed to perform the agreement (non‑performance). As a result, the limitation of the TC’s liability provided for in the terms of service would not apply.

Even if we assume that the claimant performed its obligation, but in an improper manner, it should be considered that such limitation of its liability would be unacceptable under Article 473(2) of the Polish Civil Code. The provision states that a stipulation that a debtor will not be liable for damage which the debtor may cause to a creditor intentionally is invalid. The intention, in turn, occurs where a debtor, contrary to an obligation imposed on it, acts or fails to act with the intention to cause damage to a creditor, and therefore it wants such damage to occur or accepts it when it anticipates that it may occur. Having regard to the scale of the breach of obligations on the part of the TC, the professional nature of their activity, the claimant’s knowledge on the subject matter of the agreement and its purpose, as well as the information provided to the client before the agreement was entered into, the Court decided that the TC had anticipated the possibility of causing damage to the client and it had accepted it. The correspondence between the parties clearly showed that the claimant was aware that it was the client’s intention to publish the translated book. The employees of the TC assured the client that the translation in the ‘standard’ option would be of high quality. For these reasons, the limitation of liability provided for in the terms of service was declared invalid by the Court for the purpose of this case.

In my opinion, a conclusion may be drawn from the above that making false declarations, at the stage of quoting, on the professionalism of the persons involved in the translation process and high quality of the translation output in the context of the lack of proper supervision of the translation process by the TC should be regarded as intentionally causing damage to the client.

 

Violation of the rules of the art of translation and the concept of a ‘professional translator’

Even though the profession of a translator (who is not a certified/sworn translator) has not been regulated by Polish law-makers, this does not mean, however, that each person who does a translation in exchange for money may be called a professional translator. The fact that there is no legal definition does not open the way to unrestricted discretion in construing a given concept.

In the case in question, the Court held that the assignment had been entrusted to a person who had not been a professional translator with adequate language qualifications or basic knowledge of using language-related technologies. The fact that a free-of-charge machine translation tool, namely the Google Translate tool, was used was also a proof of insufficient competence and knowledge in the area of the rules of the translation art, illustrated further, among others, by the lack of post-editing skills and ignorance of the terms and conditions of the service. Using a free-of-charge machine translation tool may constitute a ground for asserting the claimant’s infringement of intellectual property rights owned by the defendant.

The method of work applied by the translator was inconsistent with the rules of the translation art as the confidentiality of the transferred data was not ensured and intellectual property rights owned by the client/the author of the source text were infringed. Machine translation with the use of the Google Translate tool does not satisfy the condition of ‘keeping strict confidentiality’ and leads to infringing third parties’ intellectual property rights.

In my opinion, any entity providing professional specialised translation services should be aware of the risks arising from the use of tools that are not intended for professional and commercial purposes. Reading of the terms of service (in this case the Google Terms of Service for Google Translate) specified by the service provider should be considered as the absolute minimum of diligence is this regard. Far-reaching rights enjoyed by the machine translation service provider under the Terms of Service in question should result in the prohibition of using this service for translating any specialised texts protected by copyright owned by the clients.

On the professional translation market, a strong emphasis is placed on a clear indication that translations were done with the use of machine translation tools. For this reason, the two processes are covered by two different quality standards. ISO 17100:2015, which specifies procedural and quality requirements for translation services, explicitly excludes any machine translation with further post-editing from its scope. Translations done within such procedure are considered as not compliant with the quality requirements provided for by this quality standard. Machine translation and post-editing are covered by a separate quality standard – ISO 18587:2017.

The introduction to ISO 18587:2017 states:

‘(…) there is no MT system with an output which can be qualified as equal to the output of human translation and, therefore, the final quality of the translation output still depends on human translators and, for this purpose, their competence in post‑editing.’

 Importantly, the quality standard on machine translation makes a clear distinction between a ‘translator’ and a ‘post-editor’. A person who only verifies the output of the machine translation and introduces corrections is not a translator under the standard. In the case in question, 92% of the translation was the post-edited output of machine translation, while 8% was so-called raw machine translation product.

Post-editing must be agreed with the client in advance.

It should be clearly stated that the use of machine translation is certainly not entirely excluded from a professional translation process. Quite the opposite, it is a common procedure for non-literary texts of the relatively low level of complexity and difficulty, not intended for highly specialised uses, such as domestic appliance users’ guides, product descriptions at online shops’ sites, simple commercial and informative texts, simple contracts, terms and conditions of services, instructions, etc.

In the case of more complex texts, where quality requirements are higher, machine translation is used only in a supportive function as suggestions or reference materials.

It should be emphasised that in the case in question the main issue was not the very fact of using machine translation but using it in a manner that is extremely unprofessional and in breach of any universally accepted rules of the art of translation and industry best practices.

The use of professional machine translation tools is acceptable only upon the express consent of the client and it constitutes a different type of service than the traditional translation service. The machine translation service is governed by a different quality standard than the translation service (for translation service it is ISO 17100:2015, while for machine translation service it is ISO 18587:2017). In the period in which the translation at issue was done, the Google Translate tool was based on the so‑called statistical machine translation, which resulted in outputs of a significantly lower quality than those produced currently by the neural machine translation.

 As for the violation of the rules of the translation art by using the Google Translate tool, it should be explained that this claim arises from the fact that the use of any services offered by Google is equivalent to the acceptance of their Terms of Service and Privacy Policy, which raise justified doubts related to confidentiality and intellectual property rights issues. An analysis of the consequences of using Google services for professional translation purposes is included below.

 

Google Terms of Service

The Google Terms of Service[1], as applicable in the relevant period when the translation to which the judgment in question relates was performed, state as follows: When you upload or otherwise submit content to our Services, you give Google (and those we work with) a worldwide license to use, host, store, reproduce, modify, create derivative works (such as those resulting from translations, adaptations or other changes we make so that your content works better with our Services), communicate, publish, publicly perform, publicly display and distribute such content. The rights you grant in this license are for the limited purpose of operating, promoting, and improving our Services, and to develop new ones. This license continues even if you stop using our Services (for example, for a business listing you have added to Google Maps). Some Services may offer you ways to access and remove content that has been provided to that Service. Also, in some of our Services, there are terms or settings that narrow the scope of our use of the content submitted in those Services. Make sure you have the necessary rights to grant us this license for any content that you submit to our Services [own emphasis].

In the context of such terms and conditions: firstly, the translator violated the Google Terms of Service by sending to the Google Translate tool the content it had no right to use for any purpose other than translation. Secondly, by sending the full text of the book to the Google service, the translator granted Google a worldwide license to use, host, store, reproduce, modify, communicate, publish, publicly perform, publicly display and distribute the book of the claimant. It should also be taken into account that Google regulated the matter of using its services in business contexts as well: If you are using our Services on behalf of a business, that business accepts these terms. It will hold harmless and indemnify Google and its affiliates, officers, agents, and employees from any claim, suit or action arising from or related to the use of the Services or violation of these terms , including any liability or expense arising from claims, losses, damages, suits, judgments, litigation costs and attorneys’ fees. Therefore, assuming that the claimant was aware of the manner in which the service was performed, or at least it is fully responsible before the defendant for the tools used by the translator, the claimant accepted the Google Terms of Service and assumed liability for claims arising from the violation of the Google Terms of Service. According to the translator’s statement, the claimant was aware that the text was translated with the use of the Google Translate tool.

Google Privacy Policy

Google Privacy Policy also includes important points related to the use of the Google Translate tool in this case[2]. According to the document: We collect information to provide better services to all of our users – from figuring out basic stuff like which language you speak, to more complex things like which ads you’ll find most useful or the people who matter most to you online. There are two manners of collecting data. The first one is to collect data provided by the user, while the second is to collect information received while Google services are used. Google may collect data on services used by the user and on the manner in which they are used (e.g. when the user visits a site with an ad tool or when the user displays and clicks on ads and materials made available by them). Google collects information, among others, in logs: When you use our services or view content provided by Google, we may automatically collect and store certain information in server logs. Such information may include details on the manner in which the service was used, e.g. searched expressions, phone call log data, IP address, data on the functioning of the device, cookie files, etc. We use the information we collect from all of our services to provide, maintain, protect and improve them, to develop new ones, and to protect Google and our users. We also use this information to offer you tailored content – like giving you more relevant search results and ads.

 According to its Privacy Policy, Google is, therefore, entitled to collect and record in server logs any information collected from the use of Google services by the user, and in particular, it is entitled to automatically save on its servers any phrases searched by the user. When using the Google Translate service, the translator entered into the machine translation tool almost the entire content of the book, thus allowing Google to save it on its servers and to use it for its own purposes.

Finally, I would like to make several observations inspired by the ruling discussed, which may prove interesting from the perspective of daily translation and localization industry practice.

  • The fact of leaving in the edited text some untranslated words, obvious linguistic and spelling errors, as well as the presence of excerpts that are entirely incomprehensible, is equivalent to the occurrence of critical errors that make it impossible to use the translation for professional purposes.
  • In the case in question, the TC claimed that even though certain parts of the text were found to be defective in terms of language style or too literal, and their register was not appropriate for the specialised translation, the text was comprehensible, which should be regarded as enough to satisfy the quality requirement under the ‘standard’ service option provided by the TC. ‘Comprehensibility’ of the text is a minimum criterion in the case of ‘budget’ translations (with low prices and short deadlines) offered for informative (indicative) and client’s internal purposes. In such circumstances, aspects such as style, word order, vocabulary or register may indeed be seen as matters of secondary importance. In the case of translations that are supposed to be used for professional purposes, however, the fact that the text is ‘comprehensible’, while at the same time being unreadable, incomplete and abundant in linguistic defects, do not justify recognising the translation as performed properly.
  • The text intended for professional purposes must encourage the trust of the reader, also on a linguistic level. A literal approach and the use of informal expressions may result in the reader’s questioning substantive aspects of the translation, which excludes its further use or quoting. Errors being typical consequences of improper use of machine translation disqualify the text in the eyes of a professional reader.
  • It is the translation company as a party to the agreement entered into with the client that is responsible for choosing a translator. Any potential selection based on clients’ preferences should be made from the group of people properly verified by the translation company in terms of their translation competence and their ability to ensure an adequate level of translation in a given domain. The translator in this case did not satisfy such standards.
  • As for the involvement of the client in the process of translating book publications, when referring to this issue in general and without any specific reference to the case in question, it may be said that cooperation between the client/the author and the translation company/translator is highly desirable and has a positive influence on the output of the translation process. This condition, however, is not a sine qua non for the service to be provided correctly, but it is rather a good industry practice. When such cooperation is established, it is possible to agree on terminological choices, clarify potential ambiguities of the source text, and obtain reference materials and additional materials that may prove useful in the process of translating the text.
  • Bad quality of the source text has an impact on the quality of the translation. However, due diligence expected of any professional means that a client ordering the translation should be notified that the material provided is not suitable for the purpose of providing the work properly (see Article 634 of the Polish Civil Code). In this legal relationship, it is a translation company that acts as a professional that accepts an order and it should assess the material provided in terms of its suitability for the proper performance of the work. The client usually does not possess knowledge and expertise to adequately assess whether or not the material provided is suitable for translation.
  • The translation company has an obligation to provide the ongoing quality supervision and organisational supervision of the translation process and it is responsible for the selection of tools used by the translator. Similarly, the translation company must ensure quality control before the translation is handed over to the client. The fact that the service ordered did not include verification, but only editing by a native speaker, does not justify the translation falling short of basic quality standards respected in the translation industry.

The ruling described above is subject to appeal. An appeal has been filed indeed and the case will be heard by the second-instance court – the Regional Court in Poznań. I will continue to follow the case closely to share with you information on the final decision in this extremely interesting case.

[1] Google Terms of Service applicable between 1 March 2012 and 11 November 2013, available at: https://policies.google.com/terms/archive/20120301?hl=en&gl=en

[2] Google Privacy Policy applicable between 24 June 2013 and 20 December 2013 available at: https://policies.google.com/terms/archive/20120301?hl=en&gl=en

 

I think this deserves to be shared far and wide. Please give Wojciech a follow on Linkedin to get updates on the case.

 

Contact


If you would like further information about the specialised services we provide, or need a quote for a translation please feel free to contact us.

Name
Email
Message

Thank you for your message! We will get back to you as soon as possible.
There has been an eror in the form. Please validate the required fields.
© Copyright 2014 JurTrans