Anthropic warns of cyberattack risks from GLM-5.3 as debate over open models continuesMachine translation
Automatically verified and published · Generated and evidence-checked automatically; not reviewed by a human.
Anthropic argued in a report that GLM-5.3, released by Z.ai and available for anyone to download and modify, poses a major cyberattack risk. MIT Technology Review’s analysis also notes that open models can be run locally and studied by researchers, but access cannot be revoked once a model is downloaded; how to balance openness and safety remains contested.
This source permits summary display only.
Read at the original source