By Konstantinos Menyktas
Legal translation presents a challenge for both legal translators and lawyers specializing in the field. It demands a combination of practice, study, deep understanding of legal systems, extensive research, proficiency in at least two languages, and attention to detail.
Recently, in the capacity of a young legal translator, I attended the Training the Modern Legal Translator and Interpreter Conference organized by the Panhellenic Association of Professional Translators Graduates of the Ionian University (PEEMPIP) and EULITA at the Acropolis Museum in Athens, Greece.
The speeches were incredibly enlightening, offering valuable insights into the legal systems and challenges legal translators and interpreters face in both my own and other EU countries. In particular, two speakers at the conference captured my attention, as their presentations focused on the advancements of AI and legal translation. Firstly, John O’Shea, a legal translator and independent researcher, delivered an amazing speech titled “Is anyone liable for legal translation?” On the other hand, Alexandra Papoutsi, an AI, Technology, and Innovation Manager, presented a speech on the topic of “AI-generated translation: legal risks and opportunities.”
Intrigued by their insights, I delved into their published work online. Inspired by their knowledge and expertise, I decided to explore this timely and fascinating topic further. As a result, I decided to write an article that centers around the complexities and implications of AI in legal translation, aiming to shed light on both the risks and opportunities presented by this evolving technology.
Market dysfunction and AI advancements
As noted by John O’Shea’s speech, there exists a market dysfunction within the field of legal translation, often due to the urgent nature of the tasks at hand. This urgency frequently leads to assignments being delegated to bilingual lawyers or even family members, rather than to professional translators with specialized expertise. In Greece, in particular, these tasks are frequently entrusted to trainee lawyers who may have limited experience in legal terminology and only a basic knowledge of English or other relevant languages, often without compensation.
The repercussions of this approach can be significant, as it frequently results in many terminological errors and misinterpretations. Furthermore, the emergence of AI technologies has introduced a shift in the field, and, therefore, legal professionals have to adapt to new roles that involve revising the results generated by AI platforms.
However, the implications and legal liabilities stemming from such a decision pose a real challenge. As AI increasingly becomes integrated into the legal translation process, questions arise regarding accountability, accuracy, and the potential for errors that may have legal repercussions.
As machine translation and generative AI become increasingly mainstream, the risk of inaccuracies in legal translations will increase. Neural machine translation, in particular, challenges traditional notions of accuracy, with AI’s translation methods introducing a degree of fuzziness that renders inaccuracies more fluid (O’Shea & Panezi, 2023).
AI and Legal Liability
According to Sloan (2023), GPT-4 achieved a score of 297 on the bar exam in a recent experiment conducted by two law professors and two professionals from the legal technology company Casetext. This remarkable feat shows the advancing capabilities of artificial intelligence, which can now beat many law school graduates on the two-day test required for aspiring attorneys to practice law in the United States.
However, does this mean that lawyers will soon become redundant? Will we witness AI-generated machines defending cases in courtrooms and AI judges rendering verdicts? Certainly not. Such a scenario would be absurd, as numerous factors come into play during legal proceedings, including ethical considerations, emotions, deception, human error, and the nuances of witness testimony.
According to Alexandra Papoutsi, despite the remarkable advancements in Artificial Intelligence, such as its ability to generate legal documentation, offer practical utility, and even adjust its tone to convey friendliness or irony, there remain significant risks associated with AI-generated translations. These risks include a range of liabilities, including tort liability, professional obligations, contractual breaches, personal data breaches, confidentiality breaches, and other ethical and legal standards issues. Moreover, contractual agreements often go hand in hand with copyright rights, adding complexity to the question of responsibility.
So, who bears this responsibility? Is it the client, the reviser or proofreader, or the AI machine itself? One thing is certain: the machine cannot be held accountable, as explicitly stated in its usage policy, because it makes numerous mistakes. Thus, the consequences of errors can be severe, including damage to professional reputation and legal damages.
It is only humans who can assume the responsibility for ensuring accuracy and delivering high-quality results. In the event of mistakes, there must be room for negotiation and resolution to rectify errors and prevent them in the future. While errors are inevitable, when language professionals use tools, including AI, as aids, the potential for improvement is vast, making the process more efficient, cost-effective, and secure.
As O’Sheira stated in his speech, providing a service involves more than just automated translation; it requires the application of skill, competence, and expertise for the benefit of others. While automatically-generated translations may be good enough, they often fall short of meeting these fundamental requirements.
Striking the Right Balance: Integrating AI and Human Expertise in Legal Translation
In conclusion, while AI can improve the translation process, it cannot replace the critical thinking and judgment of human professionals. By leveraging AI as a tool alongside human expertise, we can strive to achieve better, more reliable, and ethically sound translations, ultimately serving the needs of clients and maintaining the integrity of legal communication.
In light of these complexities, it is important to ensure that human experts remain at the forefront of AI-assisted legal translation. When translators use machine translation tools as an aid, they typically exercise professional judgment and assume the associated risks. However, when tasked with reviewing or correcting AI-assisted translations, translators often find themselves with limited control over the specific choices made beforehand.
Therefore, I will reference Oshera’s recommendations from the conference for addressing these challenges:
Firstly, it is important to provide clear disclaimers regarding the limitations of machine translations and to encourage clients to seek professional translation and legal advice for such tasks.
Secondly, investing in higher-quality translation services is essential. This may involve integrating machine translation with human translation, review and editing processes to ensure accuracy and mitigate potential errors.
Lastly, consulting with experts in legal translation is key. Lawyers, law firms, and companies should collaborate with professionals who have specialized knowledge and experience in legal terminology and language nuances. Thus, developing a comprehensive policy for legal translations can help mitigate risks and ensure compliance with legal standards.
References:
Panezi, Argyri and O’Shea, John, How Can We Manage the Risks and Liabilities Associated With Legal Translation in the Age of Machine Translation and Generative AI? (July 01, 2023). Available at SSRN: https://ssrn.com/abstract=4707819 or http://dx.doi.org/10.2139/ssrn.4707819
Sloan, K. (2023, March 15). Bar exam score shows AI can keep up with ‘human lawyers,’ researchers say. Reuters. https://www.reuters.com/technology/bar-exam-score-shows-ai-can-keep-up-with-human-lawyers-researchers-say-2023-03-15/