Reaction
The cyber-risk trade-offs of open-weight AI models
Interconnects AI argues that restricting open-weight models while leaving closed-model APIs available could worsen cyber risks.
TLDR
In an Interconnects AI essay, the author argues that cyber policy should weigh risks from closed-model APIs alongside those from open-weight models, as well as the potential defensive uses of open weights. He says there is little public evidence of a major change in cyber harm more than a month after GLM-5.3’s model weights were released.
Combined views
1.1K
2 Sources, first seen ago
