ACM Leaders Argue It's Time to License the Digital Library for LLM Training

In a Communications of the ACM opinion piece, ACM leadership makes the case that the organization should stop holding the line and start letting large language models access its Digital Library. The framing is that AI systems have become part of the infrastructure through which scientific knowledge is discovered, synthesized, and applied — so refusing to engage risks making ACM’s corpus invisible in the tools researchers increasingly rely on. The timing follows ACM’s shift to full open access on January 1, 2026: authors now retain copyright but grant ACM a non-exclusive license to publish and to sub-license their work to third parties, explicitly including LLM training, under a choice of CC-BY or CC-BY-NC-ND terms.

The authors are careful to note this is a position statement, not a decision — no licensing deal has been signed, and they are soliciting feedback from members. They also flag real reservations: today’s AI systems rarely attribute authors or publishers properly, hallucinations can distort or misrepresent research findings, and there is no mature legal, technical, or operational framework to guarantee transparency and rights management. Protecting the integrity of the Digital Library is presented as the precondition for any deal.

The community reaction, visible in the Hacker News thread, is sharply skeptical. Critics call it hypocrisy that researchers supply the expertise for free while publishers and AI firms monetize it — in effect selling the same donated labor a second time without author consent or benefit-sharing. Others warn that pay-to-train licensing entrenches the largest AI companies while shutting out open-source models and smaller players who can’t afford access, and that gating knowledge to well-funded LLMs is a strange inversion of academia’s sharing mission. A recurring counterpoint is that these models have already ingested most of this content through other channels anyway, making the debate partly about who gets paid rather than who gets access.