SEGA Discloses Generative AI Use in Crazy Taxi: World Tour

SEGA's Steam listing for Crazy Taxi: World Tour states, "At SEGA Corporation, we utilize generative AI as a support tool for developers, aiming to provide better content to our users and enable developers to focus more on creative tasks," and adds, "We have used such generative AI support tools during development of Crazy Taxi: World Tour," with the caveat that "No AI was used in reference to the performers in the game," as reported by Eurogamer, GameSpot, and Kotaku. The game was announced at Xbox's showcase for PlayStation 5, Windows PC, Xbox Series X, and Nintendo Switch 2, with a 2027 launch window (Eurogamer). The disclosure drew immediate criticism from players and games press.

After the backlash, SEGA offered more detail, saying generative AI was used to support teams during development of the game's background assets and that the generated output remained subject to review by the development team, per GameSpot and Push Square. Push Square reports the series creator framed the use as "reference" rather than finished in-game content. The two framings sit in tension: the reference-only characterization and the description of AI-supported background-asset work leave it unclear exactly which asset categories shipped with AI involvement, and SEGA has not named the specific tools or models used. Kotaku reports the game's lead acknowledged that generative AI would remain a "hot topic" but said the team is using it regardless.

As a general pattern, studios that disclose generative AI typically cite tasks such as concepting, environment or prop iteration, localization, and text generation, and pair model-in-the-loop ideation with human curation and postprocessing to meet quality and IP requirements. Storefront disclosure itself reflects platform policy: Steam requires developers to declare AI-generated content, which is why these admissions increasingly surface on store pages rather than in marketing. The recurring friction point in comparable episodes is not the existence of AI tooling but vague disclosure, unclear crediting of human artists, and uncertainty over whether AI-generated material appears in the final build.

Watch whether SEGA names the specific tools, asset classes, or internal review steps used to validate AI-generated content, and whether store-page reviews, social sentiment, or union and artist responses escalate. Post-launch patch notes and any provenance or crediting documentation are practical indicators of how the disclosure affects studio practice. For developers and ML practitioners integrating generative systems into creative pipelines, the episode underscores the value of explicit provenance tracking, documented human-review workflows, and clear public communication when these tools touch consumer-facing assets.