Open-weights AI models may have six months before federal regulation effectively kills them. White House discussions are underway on an executive order targeting open models, with early signals pointing at Chinese-origin models like DeepSeek and government use cases first. A capability threshold is forming: any open-weights model meaningfully exceeding GPT 5.5, Claude Opus 4.8, or GLM-5.2 could trigger a mandatory government review. Given the current pace of development, that threshold gets crossed within six months.
Two policy fights are running simultaneously, and they are not the same fight. The frontier capabilities debate is driven by genuine concern about open models reaching closed-frontier performance levels. The distillation debate is something else: Anthropic has led a campaign against Chinese model makers who used its API, then recommended policy action while standing to gain directly if those competitors are banned. The author calls this regulatory capture, and makes a specific argument worth reading in full: if Anthropic cannot secure its own API, it cannot credibly claim its models are too dangerous to be replicated.
The structural problem is that open models have no single economic champion with lobbying power to match Anthropic or OpenAI. Closed model companies are easier for regulators to manage and better resourced to influence the process. Once a review framework is in place, open models will face a slower-moving threshold than closed ones. The piece argues that conceding anything on the distillation question sets a precedent that dismantles the entire emerging open-model economy, from inference companies to fine-tuning shops to downstream products.
[READ ORIGINAL →]