Crystal Dynamics explains generative AI use in Tomb Raider: Legacy of Atlantis development

Crystal Dynamics confirmed it used generative AI during the early development stages of its upcoming title, *Tomb Raider: Legacy of Atlantis*. The admission follows player backlash over the studio's Steam page disclosure, highlighting the growing friction between development teams adopting experimental tools and consumer expectations for human-crafted content.

Early-stage visualization

Jeff Adams, Experience Director at Crystal Dynamics, addressed the criticism in an interview with Polygon. He explained that the studio treats the technology as a rapid prototyping tool rather than a final asset generator. "At Crystal Dynamics, we see AI as a tool that can help our team get the right answers faster," Adams said. "In early level development, we might have the idea for an in-game object. But we might not be sure if we want to take the time to have devs build it. What we can do is use a generative AI tool to visualize it in the world. If it works, we can move it into our traditional pipeline. From there, the team can concept it and build it." Adams added that the final game assets remain human-made. "At the end, all the finished product in the final game will be human-crafted," he said.

Pipeline integration concerns

Critics argue that relying on Generative AI and LLM outputs for initial templates fundamentally alters how a project is conceptualized, even if developers replace those assets later. The initial foundation still stems from machine-generated data. This workflow mirrors broader trends where engineering teams test machine-generated concepts before committing resources. The debate centers on whether early reliance on automated tools compromises the original creative intent of a project.

Why this matters for IT and development professionals

Game studios are not the only sector testing these boundaries. Software teams across industries face similar pressure to accelerate early-stage design and prototyping. Integrating AI for IT & Development requires clear guardrails to ensure generated prototypes do not dictate final architecture or user experience. Development teams must establish transparent documentation around which pipeline stages permit automated generation and which mandate strict human oversight.