Over the last few years, artificial intelligence (AI) technologies have become part of everyday legal work, including translation. Lawyers and translators handle briefs, contracts, and regulatory filings that move across languages and jurisdictions on a routine basis, and AI-assisted translation promises to make that work faster and more efficient. However, using these tools raises critical questions concerning accuracy, confidentiality, and liability.

Accuracy and legal effect

AI-assisted legal translation refers to the use of machine translation (MT) or large-language-model (LLM) systems to generate an initial version of the text, which is then revised by a qualified human translator. These tools can significantly enhance productivity and terminological consistency, yet their output is based on statistical probability, not meaningful interpretation. Legal translation, whether AI systems are used or not, requires linguistic accuracy, but also conceptual precision and a sound understanding of the legal system underlying each language, so that legal effect is preserved across jurisdictions. At present, this lies beyond the ability of AI systems. 

Unsupervised reliance on AI systems in legal translation can have serious consequences when errors are made, as even minor lexical inaccuracies can alter obligations, liabilities, or procedural meaning. As discussed in a previous post on the implications of the new EU legal framework for AI, small changes in wording can affect how rights and obligations are interpreted, particularly as regards consent, liability, and procedural safeguards. Mistranslated privacy notices or consent terms may result in non-compliance if individuals are unable to clearly understand their rights or the scope of data processing. For this reason, oversight by expert professionals is necessary to ensure that legal effect is preserved.

Confidentiality and data governance

Confidentiality is a cornerstone of both legal and translation ethics. The EU General Data Protection Regulation (GDPR) establishes a legal framework for data protection. When a translator or company makes use of an AI system, sensitive client data may be transferred to external servers operated by third-party providers. Unless explicit contractual restrictions are in place, such data transfers can breach professional secrecy and data-protection obligations. 

As regards professional regulators, the American Bar Association (ABA) Rule 1.6 on Confidentiality of Information and its state equivalents require lawyers to safeguard client information and to use “reasonable efforts” to prevent unauthorized access or disclosure. Likewise, the Solicitors Regulation Authority (SRA) of England and Wales has warned of data-protection and confidentiality risks when law firms use external AI tools and expects solicitors to ensure compliance with UK GDPR and confidentiality obligations. Its risk-outlook materials stress that firms remain responsible for the security and integrity of client data even when using third-party technology.  

Legal practitioners and translation agencies should thus ensure that:

• AI tools are deployed within secure, closed environments;
• data are not stored or reused for model training;
• vendors comply with applicable data-protection standards; and
• translators disclose AI use transparently where required by contractual or regulatory duties.

Liability and professional accountability

Ethics experts increasingly consider AI tools as a form of “nonlawyer assistance” that falls within lawyers’ supervisory duties under rules such as the United States ABA Model Rule 5.3 on Responsibilities Regarding Nonlawyer Assistance. Comment 8 to ABA Model Rule 1.1 on competence stresses the need to understand “the benefits and risks associated with relevant technology,” which now includes AI systems used in drafting and translation. The ABA also makes it clear in its Formal Opinion on Generative AI that lawyers are responsible for work generated with these tools and must review, verify, and correct AI-assisted output before relying on it or submitting it to courts or counterparties.

UK commentary on AI in legal practice similarly emphasizes that the use of AI tools does not dilute professional obligations under SRA Standards and Regulations. Named solicitors are accountable for the quality and integrity of the work, even if AI-driven tools are involved. Recent judicial reactions to AI-generated submissions containing “hallucinated” case citations show how, when a filing misleads the court, it is the human lawyer, not the system, who bears responsibility and may face sanctions.

As for legal translation, regardless of whether AI is involved or not, responsibility for the final text continues to rest with the human translator or agency. Ethical codes, such as those of the International Federation of Translators (FIT) Translator’s Charter, specify that practitioners bear full liability for their work. The companies that provide AI models – such as OpenAI, the designers of ChatGPT – are not held responsible for their output since such models lack intent and legal standing.

Thus legal translators and law firms should implement documented quality-assurance protocols in their workflow, including:

• human post-editing and legal review of all AI-generated content;
• multi-stage validation for high-risk documents; and
• internal guidance on acceptable AI use.

Such measures uphold professional standards while mitigating risks related to negligence or misrepresentation.