Navigating ethics and copyright with AI tools

The provenance problem

The most significant friction point in adopting AI generation tools—particularly for art and audio—is the issue of provenance. Many foundational models were trained on vast datasets scraped from the internet without explicit consent from the creators. For an indie studio, this introduces both ethical discomfort and potential legal risk. You cannot confidently copyright an asset if you cannot prove ownership of its constituent parts.

Setting internal rules

At Solobit, the approach is defined by strict internal boundaries. AI is used heavily as a workflow accelerator—for coding, brainstorming, and structuring data. However, for outward-facing, final-shipping assets, the standard changes. We prioritize tools trained on public domain or explicitly licensed data, or use AI strictly as a private iteration step before a human creates the final, original work.

Transparency with players

Ultimately, player trust is paramount. If a game relies heavily on AI-generated content, that fact should be communicated honestly. Players are increasingly adept at spotting the 'uncanny valley' of unedited AI output. Building a sustainable studio means building a reputation for craft, and craft requires intentional, transparent choices about the tools you use.