ISBNdb retracts AI training offer following market scrutiny
The book database provider deleted pages offering printed texts for machine learning, characterizing the move as a test of market interest that never reached execution.
Julian Reeve
Jul 31, 2026 · 1 min read
Nine days after public attention focused on its new service offerings, ISBNdb has removed all web pages related to sourcing printed books for artificial intelligence training. The company issued a statement clarifying that it has never purchased, scanned, or sold a book for the purpose of training models, describing the recently deleted landing pages as a "test of market interest." The withdrawal marks a significant reversal for a firm that has functioned as a central catalog for libraries and bookstores for two decades.
The incident highlights the tension between legacy data aggregators and the burgeoning demand for high-quality training sets. While many AI developers seek licensed text to avoid copyright disputes, the prospect of digitizing physical libraries specifically for machine consumption remains a contentious frontier. ISBNdb's return to its core function as a metadata provider underscores the reputational risks faced by data firms attempting to pivot toward the AI supply chain without established legal frameworks or industry consensus.