There’s a perception that major AI companies have largely embraced an “ask forgiveness, not permission” approach when training their models on content scraped from the internet—including the work of countless creatives, many of whom are far from pleased (via Reuters).

Ubisoft itself has acknowledged that relying on third-party LLMs can create technical challenges. For example, the company noted that some hiccups might arise due to the recent Gemini 3 update.

“For me, we still have the choice and ability to say, ‘we want to add more emotion here,’” explained Manzanares. “It’s an iterative process. We spend a lot of time on the character sheet and content, and sometimes it just doesn’t evoke emotion.”

He emphasized that Ubisoft approaches AI experiments by starting with “something that matters.”

“It’s great to have potential prototypes, but if we don’t see the underlying code or understand how it integrates with a project or our pipelines, it’s not useful for us,” he said.

The goal, according to Manzanares, isn’t to produce flashy showreels, but to equip Ubisoft’s internal teams with tools that help them deliver engaging video games.

“It’s a modular platform because we didn’t want to be tied to a single model,” he continued. “We want the flexibility to choose whichever model fits our needs—today that might be OpenAI, Claude, or Gemini. Tomorrow it could be something else entirely, whether a new external model or one we develop in-house. If we don’t maintain that flexibility, any innovation could derail our prompt workflows and make everything fumble.”

By contrast, he believes that true impact comes when creatives collaborate closely with AI teams. “This is a real focus for me, because it’s how we can make a difference in the future,” he said. “Bringing creative talent together with AI teams lets us ask, ‘what does the future look like for players?’ We’ve only scratched the surface. There’s still much work to do with emotion and relationship systems—but we’re focused and committed.”