AI and Machine Translation policy
Introduction
Automated translation has been around for decades, but it was with the rise of Neural Machine Translation at the end of the 2010s that it began to have a major impact on the translation industry. Now generative AI, with its copyright and environmental concerns, is having an effect not only in translation, but across all sectors.
It’s undeniable that AI has many promising applications that can improve our lives, such as detecting cancer reliably. However, it is also clear that there are many things that AI doesn’t excel at, such as translating in a creative, idiomatic way that speaks to the intended audience. Moreover, AI is prone to presenting information as if it were true when it is not, which means it can’t be relied upon for research purposes.
To work, AI models require vast amounts of electricity and water. Their hardware components also require rare earth minerals, and extracting them is extremely polluting and puts peoples’ lives at risk. The safeguards in LLMs rely on the labour of underpaid workers who are forced to consume harrowing content in order to label it appropriately, which can lead to PTSD.
Moreover, most “generative” AI models have been created by scraping the internet and using copyrighted content for profit without the permission of its authors.
AI has a huge environmental and human cost that can’t be ignored in the fight against climate change and social injustice. And if that wasn’t enough, its use undermines creativity and impoverishes our languages and our knowledge of the world.
aire translations: human translations
For all the reasons outlined above, I avoid using “generative” AI and make limited use of machine translation. Aside from LLMs, AI is present in many tools, so I endeavour to use these tools carefully.
For transparency:
- I use AI-assisted digital tools such as LanguageTool (for spelling and grammar checking) and HappyScribe (for rough transcription and time-coding). Before using such tools, I make sure the data will remain confidential.
- I may use machine translation tools for specific sentences after I’ve draft-translated them myself, to check for potential synonyms.
- I may accept machine translation postediting tasks (MTPE) when the source content doesn’t have an artistic or creative intent. Research has shown that once the translator has seen the machine-translated output, it’s difficult to unsee it, and it makes it harder to come up with better, idiomatic solutions. For this reason, I don’t offer MTPE for translating films or marketing content, for example.
Read more
I wholeheartedly support the work of unions and groups, such as Arte es Ética, the Creative Rights in AI Coalition and the Por una traducción humana campaign, who are coming together to defend intellectual property rights and to tackle the damaging impact of AI on people and planet.
This policy was last updated on the 16th of December 2025.