Reddit users discuss local AI archiving strategies in light of tightening open-weight policies

On November 13, 2026, a user on the Reddit community LocalLLaMA published a post discussing how local AI developers and users should approach backing up and archiving models and training tools in case open model weights become more restricted. The poster framed the question around several developing concerns: the departure of Jensen from NVIDIA, a growing narrative that local AI is dangerous, a possible ban on Chinese models, and hosted models being locked down for reverse engineering and cybersecurity work. Against that backdrop, they said they were starting to think through a worst-case scenario in which Chinese weights are banned and US labs stop releasing open models. The post then turns directly to the community for plans, asking which models people are archiving, whether they keep only the largest and strongest ones or spread across a range, and which models they assign to specific tasks such as coding, RE/security, vision, embeddings, RAG, and small fast models for agents. It also raises the question of which training tools should be preserved. As of the available coverage, the discussion remains at the stage of an open community question, with no conclusion reached and no follow-up progress or concrete preservation efforts recorded in the post.