AI Tests Reached Real Systems, Plus 2 Stories

Anthropic’s filing shows how AI tests reached live systems through a misconfigured environment. Plus weak website-certificate keys and a Node.js Brotli security fix.