GLM-5.3's open release has yet to produce major attacks despite Anthropic's warnings about its Mythos-level cyber risk, undercutting calls to ban open models (Nathan Lambert/Interconnects AI)

Despite Anthropic's warnings about the "Mythos-level cyber risk" of GLM-5.3, its open release has not yet led to major attacks. This outcome challenges arguments for banning open models, suggesting a need to acknowledge trade-offs rather than imposing outright prohibitions. The situation highlights an ongoing debate about the safety and regulation of open-source AI.