Google Seeks Korean Book Deals as AI Copyright Bills Stall

The core issue for Korean publishers is establishing fair compensation standards for AI training data. The article highlights how a $3,000 settlement for pirated works by Anthropic is being viewed as a de facto price, rather than a damages payment. This creates a risk that an urgency-driven price accepted by one publisher could become a low market standard, making it difficult for the smaller Korean publishing houses to negotiate effectively.

The shift by global AI companies from using illegally downloaded books to direct purchase contracts introduces new complexities. Even with direct publisher deals, the absence of clear legal guidelines means publishers still face potential disputes if they sell content for AI training without explicit author consent. This highlights the need for robust internal processes within publishing houses to secure all necessary rights.

Bargaining power is uneven. Many Korean publishers are smaller than their international counterparts, and the volume of titles held by individual publishers may not be enough to counter the scale of global tech giants like Google. This imbalance is exacerbated by a lack of benchmarks for pricing and technical challenges in meeting AI companies' file format and quality requirements, further complicating negotiations for fair terms.

The industry also faces the challenge of moving beyond one-off licensing. Future models may involve AI systems referencing and citing books, requiring repeated payments based on usage. This introduces a new layer of complexity, particularly in dividing royalties between authors and publishers, a point of contention seen in some overseas settlements.