The State of GPL Propagation to AI Models
14 days ago
- #Open Source
- #AI Licensing
- #Copyright Law
- GitHub Copilot's 2021 launch raised licensing debates due to its use of Open Source code, especially concerning GPL propagation to AI models.
- By 2025, the theory of GPL propagation to AI models has waned in prominence, overshadowed by AI coding's benefits, though legal battles continue.
- Two key lawsuits, Doe v. GitHub (Copilot class action) and GEMA v. OpenAI, challenge the legality of using Open Source code in AI training without adhering to original licenses.
- The Munich I Regional Court's 2025 ruling in GEMA v. OpenAI deemed AI model 'memory' of lyrics as copyright infringement, setting a precedent for reproduction rights.
- Japanese law, under Article 30-4 of the Copyright Act, broadly permits AI training acts but leaves license compliance, like GPL propagation, unresolved.
- Arguments against GPL propagation highlight copyright law's unsuitability for AI models, GPL's unclear applicability, technical impracticalities, and policy drawbacks.
- OSI and FSF diverge on AI freedom: OSI focuses on transparency and reproducibility without mandating full training data disclosure, while FSF advocates for complete freedom, including data.
- The theory of GPL propagation to AI models remains legally uncharted, with ongoing debates on balancing software freedom with AI innovation.