The American Translators’ Association published a statement on Adoption of Technology and AI in Translation and Interpreting on 20 May 2025. The ATA’s Board of Directors approved the statement on 3 May 2025.
The statement notes that the latest wave of artificial intelligence (AI), powered by large language models (LLM), is reshaping many professions, including translation and interpreting. However, the ATA warns that growing reliance on AI highlights—not diminishes—the need for expert human linguists with specialised skills to address translation and interpreting challenges arising in this new context.
Benefits and potential of AI
The ATA points out benefits of AI in translation and interpreting:
- AI-assisted translation and interpreting has proven very useful for automating predictable, low-risk content, for drafting preliminary versions for expert revision, or in very specific scenarios. But expert human intervention is needed to discern and correct inaccuracies, inconsistencies, and misinterpretations.
- Real-world implementation of AI across translation and interpreting workflows can—under expert human oversight— lead to measurable gains in productivity, consistency, and responsiveness, particularly in time-sensitive or high-volume domains. For example, LLM-assisted workflows can accelerate terminology extraction, improve first-draft quality in repetitive or formulaic content, and enable faster multilingual content turnaround for crisis response, education, and global health.
- emerging AI-enhanced tools are helping support interpreters through real-time glossary access, pre-session briefing automation, and AI-generated transcripts that improve documentation and accessibility.
Dangers of AI
But the ATA also warns of some dangers of AI:
- AI-generated translations and interpretations may appear accurate to the general observer, making errors harder to detect for those without linguistic expertise.
- AI systems increasingly pull information from uncontrolled, unverified, biased or even manipulated sources, increasing the risk of spreading misinformation, inaccuracies, and flawed translations.
- AI translation performance varies widely across language pairs. Users may mistakenly assume that AI provides uniform quality across all languages.
- AI tools may require users to input text into platforms that may store, process, or even repurpose the data. This can lead to breaches of confidentiality agreements and of data protection laws, jeopardizing the privacy and security of the parties involved (and of the public at large).
The ATA stresses the need to use trained professionals with deep linguistic and subject matter expertise to:
- ensure accuracy and preserve intended meaning, safeguarding the interests of all parties involved, especially in high-stakes scenarios.
- ensure that AI-generated content is properly assessed, corrected, and validated.
- lead the way in shaping how AI is integrated as the broader public adopts digital-first approaches.
The ATA underlines the need for constant research and ongoing training in language and technology to augment and support the work of linguists in both traditional and new ways.
The full statement is at ATA Statement on Artificial Intelligence – American Translators Association (ATA)