AI Summary

Anthropic issued a report late last month warning that the Chinese company Z.ai’s GLM‑5.3 model presents a major cyber‑attack risk. GLM‑5.3, a frontier AI system released in August, is openly downloadable and can be modified by anyone. Anthropic’s concern revives the ongoing debate over whether such powerful models should remain open, as GLM‑5.3 does, or be restricted behind the walls of leading AI labs. The discussion highlights the tension between the benefits of open access and the potential security threats posed by unrestricted AI technology.

AI summaries can be wrong sometimes—always verify important details using the source article.