Meta legal notice forces Heretic project to remove Llama derivatives
After a Meta legal notice, the open-source Heretic project pulls Llama model derivatives.
Meta Platforms, Inc. has sent a legal notice to the Heretic Free Software Project, ordering it to remove derivatives of Meta's Llama family of AI language models. In a statement laced with sarcasm, Heretic's lead developer ('p-e-w') confirmed compliance, removing all relevant model weights from repositories it controls. The project framed the takedown as aligning 'with the agenda of the global corporate oligarchy,' while noting that Llama ranks only among the top 200 models—trailing 168 others from 23 competitors on the LM Arena leaderboard.
The irony is not lost on Heretic: Meta itself faces lawsuits and investigations worldwide over the legally and ethically dubious circumstances under which Llama models were trained. Despite the forced removal, Heretic is actively diversifying its infrastructure. The project now has an official mirror on Codeberg (hosted in Germany) and plans additional mirrors. It is also working on technological measures to ensure continued access to models created with Heretic, independent of any single service provider or jurisdiction. The move underscores growing tensions between open-source AI development and aggressive corporate IP enforcement.
- Meta sent a legal notice to the Heretic project over derivatives of Llama AI models, prompting a sarcastic but compliant removal of model weights.
- Heretic is now mirroring its project on Codeberg in Germany and implementing tech measures to avoid future dependency on US-based services.
- Meta itself faces multiple lawsuits regarding Llama's training data, highlighting the irony of its aggressive IP enforcement.
Why It Matters
Highlights the clash between open-source AI innovation and corporate IP enforcement, with implications for model availability worldwide.