Anthropic's latest Claude Code release lets a running session fork into a parallel branch, adds gateway request headers and reconnection notifications for MCP servers, and fixes several permission-checker bugs.
why it mattersSession forking and clearer MCP reconnection state target people running longer, more complex agent sessions.
Anthropic (Claude Code changelog)
Security researchers reported that a Russian-speaking actor orchestrated hundreds of autonomous AI agents, built on OpenAI Codex and a DeepSeek model, to exploit two PaperCut NG/MF vulnerabilities, compromising over 440 print-server instances across 395 organizations in 48 countries, with the fastest breach taking under four hours from a standing start.
why it mattersOne of the starkest public examples yet of coding agents being repurposed for automated, large-scale exploitation rather than development.
The Hacker News