FSF Threatens Anthropic Over Alleged Copyright Infringement of Shared LLMs

The Free Software Foundation’s warning to Anthropic marks the first high‑profile legal challenge over the distribution of shared large‑language models (LLMs). In a post on its licensing blog, the FSF alleges that Anthropic’s “Share‑Your‑LLM” initiative violates the GNU General Public License by bundling code that incorporates GPL‑covered components without providing the requisite source code and downstream rights (FSF blog). The foundation says the breach is “material” and threatens to file a lawsuit unless Anthropic complies within 30 days, a timeline that mirrors the FSF’s typical enforcement cadence for open‑source violations.

Anthropic’s response, while brief, underscores the company’s reliance on a settlement framework it introduced last year to encourage open‑source contributions to its models. The FSF’s filing, however, points to a specific version of the model that the foundation claims was derived from a GPL‑licensed training pipeline. According to the FSF, the model’s weights and accompanying scripts were made publicly downloadable on Anthropic’s GitHub repository, yet the accompanying source for the underlying training code was omitted, contravening the license’s copyleft provisions (FSF blog). The foundation’s legal team has reportedly prepared a cease‑and‑desist letter that also demands retroactive compliance for all prior releases.

The dispute has quickly become a rallying point within the AI community. A petition circulated among engineers at OpenAI and Google, highlighted by The Verge, urges the companies to “stand with the FSF” and pressure Anthropic to honor open‑source licenses (The Verge). The petition’s signatories argue that lax licensing practices could erode trust in collaborative AI development and set a dangerous precedent for future model sharing. Meanwhile, TechCrunch reported that several tech workers have called on the Department of Defense and Congress to reconsider any contracts that label Anthropic as a “trusted vendor” until the licensing issue is resolved (TechCrunch). Those critics note that government procurement processes often lack rigorous checks on software licensing compliance, potentially exposing taxpayers to legal risk.

Reuters’ coverage of Anthropic’s broader business strategy adds context to the timing of the FSF’s threat. In a recent breaking‑views piece, the outlet described Anthropic’s aggressive revenue forecasts as “hallucinations” that rely heavily on the company’s ability to market proprietary versions of its models while simultaneously courting the open‑source community (Reuters). The article suggests that the FSF’s action could undermine Anthropic’s narrative of openness and may force the startup to recalibrate its product roadmap. If the lawsuit proceeds, Anthropic could face not only legal costs but also a reputational blow that might deter enterprise customers wary of licensing entanglements.

Legal scholars familiar with software licensing note that the FSF’s move is consistent with its historical enforcement of the GPL, which has included actions against major corporations such as Samsung and Cisco. The foundation’s counsel, quoted in the blog, warned that “non‑compliance undermines the very ethos of free software and threatens the collaborative ecosystem that fuels AI innovation.” Should the case go to court, it could set a precedent for how AI model distributions are treated under existing open‑source licenses—a question that has remained largely untested despite the rapid proliferation of LLMs. For now, Anthropic faces a tight deadline to either amend its distribution practices or risk a protracted legal battle that could ripple across the AI industry.