Monash Scholars Argue AI Distillation May Not Infringe Copyright

Legal scholars at Monash University have published an analysis arguing that AI model distillation likely does not constitute copyright infringement under current law. The paper contends that extracting functional capabilities from teacher models like Claude or GPT represents learning unprotected ideas rather than copying protected expression. This distinction challenges lawsuits alleging that training on synthetic outputs violates intellectual property rights. The authors assert that copyright protects specific creative expression, not the underlying knowledge or utility embedded within model weights. While rights holders claim distillation amounts to high-tech piracy, this academic framework suggests existing legal doctrines already accommodate technical learning processes. The analysis arrives amid escalating litigation between AI laboratories and content creators regarding training data usage. If courts adopt this reasoning, it could significantly narrow the scope of viable copyright claims against foundation model providers utilizing synthetic data pipelines.