Big Walk: Using an LLM to Fix a Bug Divides IGF Jurors

House House has acknowledged using a chatbot to resolve a Unity bug in Big Walk, affecting less than 0.1% of the code and resulting in ten modified lines. This revelation, deemed late by some, has triggered an organized reaction from IGF jurors and a wave of negative reviews, while others call for a distinction between technical assistance and a massive reliance on AI.

Canceled Awards and a Wave of Negative Reviews Revive the Hardline Stance

Last year, the Indie Game Awards stripped Clair Obscur: Expedition 33 of its accolades after the admission of using AI tools for certain textures, which were later corrected. The organizers maintain a firm stance against generative AI during nominations and the ceremony. Since House House's disclosure, jurors in favor of excluding AI have received broad support from disappointed fans and observers. On Steam, Big Walk accumulated over 400 negative reviews in just a few days, although anti-bombing mechanisms seem to limit the impact on the overall rating, which remains labeled as Very Positive. For many, a strict public position remains useful in the face of a phenomenon perceived as detrimental to the industry.

Targeted Use of LLMs to Resolve a Unity Bug

House House indicates that it used LLMs on a one-off basis towards the end of production to address technical lockouts. A problem with the Unity engine was causing shadows of infinite length, slowing down the game in some cases. After a Unity engineer determined that no fix was possible, the team consulted a chatbot that suggested a lead, which was then extensively reworked before being integrated into ten lines of code. The studio estimates this contribution at well under 0.1% of the project. However, one internet user argues that a human fix was still feasible at the cost of significant effort.

Disclosing Where, When, and How: Valve, IGF, and the Scope of AI

When releasing on Steam earlier this year, House House did not detail this use of a chatbot, believing that Valve does not require code disclosure or mention of chatbots. The platform's form targets AI-generated content and not efficiency gains from tools integrated into development environments. In contrast, the IGF does not separate these cases, which led House House to report the AI-assisted correction in its application. A version of this note was later made public, with the studio deeming the issue of public interest.

A Collective Stance from IGF Jurors to Ban Generative AI

Nineteen jurors, out of a panel of over fifty, announced they would refuse to vote for any game incorporating generative AI, citing an increase in relevant submissions. They mention ecological, economic, psychological, and social damages related to labor and remind that AI is not a given for the sector. The group calls on the IGF to adopt a strict policy prohibiting any AI-generated content in the future.

Between Admiration for the Game and a Break on Principle

Jurors who had praised Big Walk for its qualities have shifted to a principled opposition. Austin Walker described the game as a remarkable achievement, while Damon Reece referred to it as a rare emotional experience. However, both are among those for whom even a reduced use of AI nullifies their support for the work in the context of awards. Reece calls for embracing the pain of a choice aimed at prioritizing human labor and establishing a line deemed necessary.

Transparency, Gradation of Uses, and Fear of Perverse Effects

House House claims it does not advocate for the use of AI tools and expresses concern that voluntary admissions could normalize the technology. The studio supports those seeking to limit the damage and apologizes to those whose standards it may not meet. Unlike zero-tolerance positions, other actors demand a nuanced reading: distinguishing a one-off technical assistance from massive use for art, music, or engines. IGF judge Liz Ryerson does not consider the matter disqualifying and worries about increased punishment for transparency. Curator Dominic Tarason describes art that is at worst very slightly compromised, recalling the years of work in contrast to a brief solicitation of a chatbot. Critics argue that preferring AI over hiring an expert dehumanizes the process; researcher Mike Cook sees this as a symptom of a larger issue. For Brendan Keogh, absolute inflexibility is counterproductive and drives some to hide more problematic uses. As engines increasingly integrate AI, avoiding any contact with generated fragments could become impractical. Conversely, Hannah Nicklin defends zero tolerance as a political stance aimed at fostering solidarity around reducing harm.