Sinofsky Defends AI Training on Discarded Books

Steven Sinofsky retweeted commentary noting that hundreds of thousands of tons of mostly obscure and out-of-print books are pulped or landfilled annually because copyright rules block digitization. He and others contend that scanning the volumes for AI model training is preferable to outright disposal. Critics including Taylor Lorenz and Brian Roemmele condemn the physical destruction of copies by companies such as Anthropic, attributing the practice to copyright constraints and calling for an end to spine-cutting and pulping. The discussion highlights ongoing tension between data needs for large language models and preservation concerns.