Industry
LEVEL-5: No Direct AI in Games, Vision PVs Being Corrected
The company is narrowing its use of generative AI after new criticism and will rework faulty presentation videos.
LEVEL-5 says that generative AI is not used directly in the actual games and wants to rework parts of the presentation "LEVEL5 VISION 2026 II 夢" after recent criticism. At the same time, the company is drawing a firmer line on AI use for future publicly shown videos.
In a statement published on September 17, LEVEL-5 announces that individual PVs from the presentation will be re-uploaded in corrected versions ahead of the start of Tokyo Game Show 2026. The aim is to fix errors caused by AI use, as well as passages that looked like AI output because of human mistakes.
The company goes into particular detail on the material for "Yo-kai Watch 2 Hado." According to LEVEL-5, several details criticized as possible AI errors were in fact deliberate simplifications or mistakes in manual work. At the same time, the company acknowledges that a too-careless approach to AI in parts of the presentation damaged trust in the entire production process.
LEVEL-5 Sets a New Boundary for User-Facing Material
The key point is the new wording on the actual development process: LEVEL-5 explicitly states that AI is not used directly in the game content itself. In the future, generative AI is to remain limited to tasks aimed at increasing efficiency. For videos and other material published directly to users, the company wants to avoid direct use.
The new statement thus supplements the apology from Akihiro Hino on September 12, which had initially focused more on the distinction between technical support and human creative work. Hino had said that scenarios, character designs and basic concepts should continue to be developed by people.
The debate is part of a broader discussion about generative AI in Japan's creative industry. For "Seishun no Yarinaoshi wa Isekai de", individual shots are generated with generative AI and then edited; a competition involving Yuki Kaji's voice AI, by contrast, explicitly pairs synthesized voice with human voice acting. The case of an AI graphic using a ShiroManta illustration also shows that handling generated content now affects several areas of Japan's creative scene.
With the new explanation, LEVEL-5 is for the first time stating more concretely where it intends to distinguish in future between internal efficiency gains and published final material.