Lost in AI Translation: Hellebore’s Stance on Generative AI
Like many folks in many disciplines these days, Hellebore finds generative AI to hold a lot of negative potential. The publishing world is already deep in the trenches, fending off waves of AI-generated manuscripts, cover art, and marketing materials. Some corners of the industry seem a little too eager to embrace these shortcuts, but most others are drawing firm boundaries. At Hellebore, we fall into the latter.
Hellebore has taken a clear position: we do not accept AI-generated writing in our catalog. This is reflected in our contracts, as it has been in the contracts of several presses we’ve previously worked with. That said, we want to be transparent about what we do allow.
We recognize that AI can function as a tool. Because of this, we’re not opposed to authors using large language models (LLMs) for research, brainstorming, or outlining. This can speed up preliminary work, such as summarizing historical events, generating comparative notes, or assisting with early market research.
However, using AI to write any portion of a manuscript is not permitted. The core voice, ideas, and prose of a book must remain the author’s own. And NEVER put any parts of your original writing into an LLM. Once an LLM learns information, it may memorize it and use it to generate future responses for anyone else using the tool, which is why so many authors have been seeing copycats of their work.
Our AI rule extends to our translation policies. For overseas editions, we stipulate that AI not be used to translate our authors’ work. Translation is one of the most nuanced, intricate, and profoundly human creative acts. It requires the translator to understand not only the words on the page, but the tone, cadence, cultural weight, and emotional undercurrents of the original text. We wholeheartedly support the skilled labor of translators and will continue to protect the integrity of their craft.
Even outside of publishing, audiences seem to have a consistent and mostly negative reaction to AI-generated content, often dubbed “AI slop.” People have learned to look for and recognize the inconsistencies, distortions, or just all-around bizarness of auto-generated art.
Recent examples abound. Coca-Cola faced criticism for a holiday advertisement riddled with odd inconsistencies. Taylor Swift was accused of using AI imagery in promotional videos for her latest album. In publishing, the backlash has been just as immediate. The cover art for Quicksilver became a flashpoint when readers noticed telltale signs of AI generation—which the author denied—but the original cover was still widely mocked as looking cheap and uncanny. A new cover was shortly rolled out, seeming to be AI-free.
Closer to home, Powell’s Books drew public ire after releasing merchandise that appeared to rely heavily on AI-assisted art. ILWU Local 5, the union representing Powell’s workers, reported that in-house designers had raised concerns before production but were ignored. Online, many criticized Powell’s for bypassing local artists in favor of low-quality AI imagery. When Powell’s did respond, they attempted to downplay the issue, stating that a “local artist” had used Adobe tools with “some AI-assisted features.”
This type of fallout really is interesting, especially when you consider that AI has infiltrated just about every sphere. Even still, we as humans seem to continue to reject AI art when it feels cheap, lazy, or disrespectful to human labor and creativity. Hopefully, everyone from executives to artists will continue to ask themselves what really matters when it comes to production and consumption. Quality and integrity do still count, regardless of how technology might evolve over the coming decades. ✿
