Anthropic recently released a report warning that GLM-5.3, an open model from the Chinese company Z.ai, poses a serious cyberattack risk because anyone can download and modify it. That claim is the latest flashpoint in a long-running argument over whether frontier AI models should be freely available or kept behind the walls of a few major labs.
Supporters of open models point to practical benefits. They can be run locally, which helps businesses protect sensitive data and reduces costs. Researchers can inspect their inner workings, trace their reasoning, and fine-tune them for new tasks. More broadly, open models are seen as a way to prevent a handful of companies or governments from controlling a foundational technology.
The trade-off is real. Closed models are currently more capable, though open ones are only months behind. Commercial incentives also shape the debate: Nvidia's CEO favors open models, while Anthropic's CEO urges caution. As one researcher put it, how you come down on the openness debate often depends on how you come down on the AI debate itself. Any policy approach, the article concludes, involves balancing safety against democratization.We need to write a short article for