Hackers Found Inside Claude Code. Anthropic Has Informed Australian Enterprises.

Hackers accessed Anthropic's development environment. The environment contained the code for Claude's safety mechanisms. Australian enterprises were notified. The original safety protocols did not account for compromise at the source layer.

Security checkpoints typically assume the upstream is secure. When the upstream is the safety layer itself, you have a structural problem that looks like a procedural one. The department that failed to include this scenario has been eliminated. Accountability has been satisfied. The vulnerabilities they missed remain in systems handling australian enterprise data.

Anthropause assumes its safety layer was not successfully exfiltrated. It has no way to know. Competitors now understand the architecture they were trying to protect.