AI Giants Push Back Against Restrictions on Open-Weight Models

The signatories argue that policymakers must distinguish between unlawful misappropriation and standard industry practices. At the heart of the debate is distillation—the process of using one model’s output to refine another—which the companies describe as a foundational technique for innovation. By framing distillation as a legitimate research tool rather than a security threat, the group hopes to prevent the White House from enacting broad bans that could cripple the open-source ecosystem.

This pushback highlights a deepening rift within the AI sector. While companies like OpenAI and Anthropic advocate for tighter controls to protect their proprietary interests, the signatories contend that open-weight models are essential for cybersecurity. They point to instances where closed-source models failed to assist in defensive tasks due to rigid guardrails, forcing researchers to turn to accessible, powerful alternatives like the Z.ai GLM 5.2 model to identify and remediate vulnerabilities. Infrastructure providers like Microsoft and Nvidia also maintain a clear economic stake in this outcome; as models become more commoditized and interchangeable, demand for cloud capacity and hardware processing power accelerates, cementing their role in the broader AI supply chain.