OpenAI Feared "Optics" of what might appear on Hacker News

Unsealed legal briefs in Authors Guild v. OpenAI allege that OpenAI and Microsoft executives and engineers knowingly trained large language models on copyrighted books obtained from piracy sites like LibGen. Internal messages show awareness that the models would replace human writers (“will make people unemployed”), efforts to tune models to “autocomplete” other authors’ work (including an explicit aim to finish George R.R. Martin’s series), and a dismissal of authors’ losses as “acceptable economic disruption” likely to produce the “death of the reader.” Executives framed their worry as an image problem - fearing headlines that OpenAI “uses copyrighted data from sketchy russian website” on Hacker News - rather than a legal or ethical barrier, and later initiated “Project Clear” to delete LibGen files and erase evidence.

Those documents underpin a class-action claim brought by the Authors Guild and dozens of prominent writers, who argue the admissions prove deliberate infringement and existential harm to the publishing ecosystem. The filings cited (Class Plaintiffs’ Memorandum, Docket 1982; Corrected Rule 56.1, Docket 1987) situate Microsoft and OpenAI’s knowledge at the center of Alter v. OpenAI and Microsoft, with Susman Godfrey’s Justin A. Nelson leading the plaintiffs; further briefing and a Manhattan hearing are expected in early 2027.