Kotaku reports, citing Bloomberg, that developers on Epic Games’ unannounced Marvel and Star Wars shooter were told to use Anthropic’s Claude in an “AI-first” production model. The report says handwritten code was no longer to be used and that art should be generated through full AI workflows. Epic communications director Jake Jones said the company was embracing AI-accelerated development while Epic developers remained the project’s creative drivers. These are the observed facts available in the supplied source; the alleged internal directive is attributed reporting, not independently verified here.
Mechanism: When a production mandate spans both code and art, AI output enters multiple workstreams, increasing the need for explicit review ownership and criteria for accepting, revising, or rejecting generated work. Kotaku says the guardrails for fixing mistakes and keeping unwanted output from the final product were unclear. The source does not document actual workflow controls, quality outcomes, or how much work Claude has produced.
Why now: The report describes a studio applying an “AI-first” instruction to a named, high-profile game project, making adoption policy—not just individual experimentation—the relevant change. First implication: If the reported scope is accurate, verification and creative review become operational dependencies of the production model, not optional cleanup.
Evidence boundary: The directive and details about staff response come through Kotaku’s account of Bloomberg’s reporting; Epic’s quoted statement confirms an AI-acceleration posture but does not verify every reported instruction. First watch condition: Look for Epic to clarify the directive’s scope, human review responsibilities, and rules for generated code and art.