At a recent publishing workshop, an editor reported receiving a literary translation that still contained the line "Pay for ChatGPT premium version!". The account did not specify whether the translator had used artificial intelligence (AI) to render the entire book or had only fed certain passages into a machine, but the incident illustrates the temptation large language models pose within the profession.
According to literary translators interviewed by HVG, many practitioners have in recent years begun to cooperate with algorithms: translator tools such as DeepL and large language models like ChatGPT have become routine aids for some. At the same time, some translators seek to exclude these tools from their workflows entirely.
Dányi Dániel, who translates between Hungarian and English and serves on the board of the Magyar Műfordítók Egyesületének (MEGY), says he deliberately avoids tools that might influence the decisions he makes during translation. For him, this kind of "AI abstinence" is a defensive tactic.
Tótfalusi Ágnes, who has translated works by Michel Houellebecq and Stendhal from French into Hungarian, likewise rejects machine assistance. She argues that AI does not meaningfully speed up the translation of easy texts and actually muddles more difficult passages; she also enjoys playing with complex sentences and feels that machine tools would spoil that pleasure.
Translators who do use AI at least occasionally generally do not upload entire manuscripts to these services. Most treat the technology as a digital assistant: for checking, exploring alternative phrasings, or testing solutions to specific translation problems. The conversations gathered by HVG suggest that only a minority try to remove AI completely from their professional practice.
The debate highlights that in the translation field AI is viewed both as a technological opportunity and as a potential erosion of creative control. Division within the profession and the concrete ways in which AI is applied are likely to continue evolving as large language models develop and become more widely available.
Issues to watch going forward
- Copyright and data protection: sending full texts to machine services can raise legal and data-handling concerns.
- Quality control: machine-generated suggestions require human review, especially for literary texts.
- Professional autonomy: some translators worry that AI may influence stylistic decisions and the creative process.
Continued discussion within the translation community and sharing of practices should help the profession define responsible and professional frameworks for using artificial intelligence.
