The Pros and Cons of Using AI for Multinational Cases | Chapter 3

The Potential of GenAI in E-Discovery Translation

AI and GenAI present massive opportunities in litigation, as well as several new challenges.

Understanding where and how different AI technologies can help or hinder eDiscovery is essential.

Machine translation has become more common in e-discovery in recent years because of its efficiency.

GenAI has the potential to be a very powerful e-discovery translation tool for cross-border litigation.

Many are turning to a hybrid approach that combines AI and human intervention to improve the efficiency of e-discovery.

AI is here to stay. However, legal teams must still include sector-specific linguists to handle nuanced translations.


GenAI, however, has the potential to be a very powerful e-discovery translation tool for cross-border litigation because it doesn’t just spit out literal translations like machine translation.

The Potential of GenAI in E-Discovery

When properly trained and maintained, LLMs built on neural networks provide a more nuanced linguistic lens, bringing in more context and idiomatic insights into the translation––and learning over time––so that the document doesn’t just say the same thing, it means the same thing in both languages.

For example, when it comes to patent litigation across borders, which is a highly technical field demanding knowledge of international patent laws and classifications, emerging LLMs can look at a portfolio of patents to build a body of context and language, avoid potential drafting errors in the future, and manage patent case flow, as well as provide more situationally specific translations. Legal teams that use GenAI over multiple legal matters will have a robust, centralized database to improve translations over time.

These positive qualities are enhanced by GenAI’s user-friendly UI. With most large providers, users simply need to enter a prompt stating what language is needed and what the translational parameters are in order to get the desired result. As legal organizations begin to build proprietary GenAI systems trained on bespoke datasets with custom UI to match, these e-discovery processes will be even easier.

But, again, this is nowhere near a perfect science yet. GenAI is still largely an experiment; off-the-shelf AI often misses the mark on cultural and contextual interpretations and regularly spits out completely false answers.

Training a bespoke model is not always attainable, and it poses major data privacy risks when sensitive information is involved. For all of these reasons, GenAI translations are not yet defensible in court and can’t be ISO-certified as meeting the highest standards.

Further Reading: Navigating ISOs – A Guide to Selecting a Language Service Provider

Search