Preserving balance in the EU digital single market: How <i>Like Company</i> could reframe copyright and innovation in the generative AI era

Abstract This study analyses the Court of Justice of the European Union's first referral on generative AI and copyright, that is, Case C‐250/25 Like Company v. Google, which interrogates whether the generation of chatbot summaries of press articles infringes the press publishers' right under Article 15 CDSM Directive and the reproduction and communication rights under Articles 2 and 3 InfoSoc Directive. The analysis shows that large language model training cannot, as such, be characterised as reproduction because it encodes statistical patterns rather than fixing protected expression, while recognising that further empirical evidence would be valuable to clarify the precise technical steps involved in the training pipeline. RAG‐generated summaries that paraphrase content within ‘very short extracts’ fall outside Article 15(1) and do not communicate publications to a new public. Alternatively, even if training were characterised as engaging temporary reproductions, the text‐and‐data mining exception under Article 4 CDSM Directive lawfully authorises such uses absent valid opt‐outs. Regarding liability, the study concludes that individual users, not platform providers, bear responsibility for outputs generated through their models' learned parameters and retrieval mechanisms. This comprehensive analysis ensures coherence across EU copyright doctrine and sustains legal certainty for AI innovation within the EU digital single market.