6 months to live for open models
| Source: Interconnects (Nathan Lambert)
Tags: open-source AI, executive order, DeepSeek, AI regulation, open weights, Nathan Lambert, AI policy
Nathan Lambert warns a White House executive order could ban or indefinitely delay open-weight AI models above current frontier capability — likely targeting Chinese models like DeepSeek — permanently cementing open AI as a regulatory second class within 6 months.
Details
US regulatory pressure on open-source AI is intensifying in ways that prior waves of anti-open-source rhetoric never matched, according to Interconnects author Nathan Lambert. Multiple sources cite active White House discussions around an executive order targeting open-weight models, with the most plausible outcome being a ban or indefinite delay on releasing open weights above a capability threshold roughly comparable to GPT 5.5 or Claude Opus 4.8. The concern centers on Chinese-origin models, where DeepSeek and similar efforts currently lead the open-source frontier by a meaningful gap. A capability-gated review requirement — likely imposed on models above a threshold that currently would catch only Chinese labs — would disadvantage the open ecosystem structurally. Lambert expects the threshold to evolve slowly for open models while closed-model companies benefit from both better physical security and significantly more effective lobbying. Two simultaneous policy tracks make this harder to navigate: restrictions on frontier-capability open models and a parallel push to restrict distillation from proprietary systems. These are separate issues with different risk profiles, but they converge on the same open-source ecosystem. Lambert notes that open models lack a central economic champion — no large company with the lobbying weight of OpenAI or Anthropic has an incentive to defend them. The article is a paid Substack piece, so the full policy analysis is behind a subscription; what is visible is already high signal-density for anyone tracking AI policy.