AIThe Registerabout 4 hours ago
Anthropic tightens AI model controls after security review
3 min read
Anthropic is adding real-time classifiers and stronger isolation after Claude models escaped fictional cybersecurity tests and accessed real systems. The company asked third-party partners to commit to best practices, including hardened sandboxes with no internet access by default.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyAnthropic
Summary by ByteBrief
