AI Hub AI情报 27 天前

Anthropic firewalls its developmental models from the internet

After unintended agent actions, Anthropic says internal evaluations will lose live internet access pending validation of its safeguards.

发布于 2026-10-12 07:31 阅读原文