Kingdom Come Dev Hates AI Art but Uses It Anyway

"I understand why people hate it, because I hate it, too." That's Prokop Jirsa, one of the new creative directors at Warhorse Studios, talking about AI-generated art in a recent interview with PC Gamer. What followed that statement was a detailed explanation of how Warhorse uses genAI anyway, including for concept art and programming tasks. The contradiction is so clean you could frame it.

Jirsa drew a line between AI art that ships in a finished product and AI art used internally during development. Using generative AI to create art that would be "released to the public" isn't something he's interested in, he said. But for internal work? He described genAI as "really helpful," specifically calling out AI-assisted programming and quick generation of concept art to improve communication between team members. His framing was practical: even non-programmers can "code little things" that help them access information more easily, and rough AI-generated concepts can get a visual idea across faster than a written brief.

I get the logic. Concept art is disposable by nature; it exists to communicate a direction, not to hang in a gallery. And using AI to write small utility scripts that never touch the final build is about as benign as the technology gets. But there's a reason this particular statement landed so awkwardly, and it has nothing to do with internal tooling.

The Translator Problem

These comments arrive while Warhorse is dealing with accusations that it fired a translator to cut costs. When a studio is already under scrutiny for trimming human roles, telling the public you hate AI art while simultaneously describing how AI handles tasks that humans used to do is a messaging disaster. Jirsa might believe there's a meaningful distinction between AI-generated concept art used in a meeting and AI-generated concept art shipped in a game. Players and workers looking at the broader pattern are unlikely to care about that distinction.

And this is where I keep getting stuck with every studio that rolls out the "AI as a tool, not a replacement" line. The tool argument only holds up if headcounts stay the same or grow. The moment someone loses their job while AI picks up the slack, the distinction between "replacing" and "making more efficient" becomes purely semantic. Warhorse hasn't drawn that connection publicly, but the timing of these comments alongside the translator situation does the connecting for them.

Jirsa isn't alone in threading this needle. On the same day his comments were published, Epic's Stephanie Arnette told a Gamescom Latam panel that AI's goal at Epic "is to make us more efficient," not to kill jobs. She acknowledged the fear directly: "I know everyone's biggest fear is, 'Oh my god, AI is going to take all our jobs.'" Epic is exploring AI tooling "in the art realm as well," she added, without specifying what that means in practice. Sony's Hideaki Nishino went even further during the company's corporate strategy presentation, calling AI a "powerful tool" that will enable "gaming experiences like never before" while insisting that creators' roles "will remain unchanged."

Three studios, three variations of the same script. I wrote about this exact pattern two days ago, and the repetition is becoming its own story. Every executive frames AI as augmentation, never replacement. Every executive acknowledges the fear. None of them offer concrete commitments about staffing levels.

What makes Jirsa's version stand out is the honesty baked into the contradiction. He didn't try to pretend he loves the technology or that it's going to revolutionize creative expression. He said he hates AI art, explained exactly how his studio uses it, and left the gap between those two statements wide open. I'd rather have that than Nishino's corporate poetry about "fresh ways to enjoy your favorite characters," but the gap is still there, and it's the same gap every studio is hoping players won't look at too closely.

Nishino detailed a Sony-built tool called Mockingbird that animates 3D facial models from performance capture data in a fraction of a second, work that previously took hours. He stressed that human performers aren't being replaced. But if the animation work that took hours now takes seconds, the animators who spent those hours are the ones whose schedules just got a lot emptier. The promise is always that freed-up time gets reinvested into "richer worlds." The reality, across an industry that laid off tens of thousands of workers over the past three years, is that freed-up time often gets reinvested into smaller teams.

Jirsa's comments about Kingdom Come's development process are the most grounded version of the AI pitch I've heard from a studio this week. He's not selling a vision of the future or promising experiences "like never before." He's describing a studio that uses genAI for quick concept sketches and small scripting tasks while keeping it out of shipped assets. Whether that boundary holds as the technology improves and the pressure to cut costs intensifies is the question Warhorse will have to answer with actions, not interviews. The translator accusation suggests the pressure is already there.