Some IGF Judges Will Refuse to Vote for Games Made with AI — Even When It Accounts for Less Than 0.1% of the Code

You spend several days wrestling with an engine bug, ask an engineer you know, and read the documentation. Then a chatbot suggests a code snippet, you rework it, and the problem goes away. For a developer, this is an ordinary story of hunting for a solution. For some Independent Games Festival judges, such an episode is now sufficient reason not to vote for a game. This is exactly the boundary at the center of the dispute involving Big Walk from House House, the creators of Untitled Goose Game.

The practical question here is quite specific: what consequences can the use of AI in development have, even if the player never sees a single generated image or line of dialogue? The story shows that a platform's requirements, a festival jury's convictions, and a studio's internal rules can diverge. It is easier to sort this out before submitting an entry, while the team can still agree on a unified approach.

What Happened with Big Walk

According to The Verge, some IGF judges published a statement saying they would refuse to vote for games in whose creation generative AI was used in any capacity. Among the signatories are Sam Barlow, creator of Telling Lies and Immortality, and Lena Raine, composer of Celeste. The authors of the statement specifically clarified that they do not speak on behalf of the festival or all members of the jury.

This is a significant caveat. Nothing described amounts to an official IGF ban. This is the position of specific judges, which may affect their votes. According to them, the growing number of submissions that use generative AI has provoked heated debate, and there is no general consensus within the jury.

House House, for its part, disclosed the circumstances of Big Walk's development. The studio says it used an LLM occasionally in the final stage of work, mainly to investigate problems with third-party code. At times, fragments of responses became a starting point for its own code after refinement. By the team's estimate, significantly less than 0.1% of the game's code is related to this. For other development tasks, the studio says, generative AI was not used.

Why Fixing a Bug Also Became a Point of Contention

One of House House's examples is a problem with shadows. The team describes spending a long time investigating it and even talking to a Unity engineer before turning to a chatbot. The developers modified the code proposed by the model and used it to solve the problem.

Imagine a similar situation on your own team: a library renders an object incorrectly, searching doesn't help, and a model suggests which part of the implementation to check and rewrite. The final solution passes through a developer, but the LLM's involvement remains in the history of how it came about. For the judges who signed the statement, that is enough: they explicitly classify generated code as the kind of AI use they oppose.

That is why the dispute is not reducible to a percentage of lines. A small share of code does not answer the principled objection of people who reject the very way it was obtained. At the same time, House House's estimate remains the studio's own claim: the material provided contains no independent verification of this figure or a description of how it was calculated.

What to Decide Within Your Team

House House said it disclosed its use of AI in its IGF submission. However, there is no such disclosure on Big Walk's Steam page: according to the studio, the platform does not ask for information about the use of chatbots or code in the circumstances described. This is House House's explanation, and turning it into a universal guide for publishing other games would be a mistake.

For a founder or product manager, a useful takeaway is to define acceptable scenarios in advance. Can you ask a model about documentation? Take a suggested snippet as a basis? Who checks the result and records the origin of the solution? Then these rules need to be compared against the requirements of specific platforms and competitions.

House House itself acknowledged the lack of an AI usage policy as a mistake and promised to develop fuller rules. Such an agreement does not guarantee approval from every jury. But it helps a team make decisions consistently and explain them when a technical way of fixing a bug becomes a public question of principle.