开源模型安全威胁将至
AI 导读
未来一段时间对更广泛的软件行业来说会很有意思，它们不得不迎头赶上，以期达到新的平衡——放弃对漏洞的懈怠态度。 让我们看看开源网络模型公开几个月后，felonybench 的分数会是什么样。
正文
Would be interesting days ahead for the broader software industry, who has to catchup to hopefully a new equilibrium -- abandoning its complacency towards vulnerabilities.
Let's see how felonybench scores look after a few months of oss cyber models being out in the open.
Leaving aside Anthropic's incentives for publishing this research, there is no doubt that open weights models will soon create the same security threats that closed source models have been demonstrating, except without guardrails. We are close. Probably good to plan accordingly.在 X 查看被引用的帖子