Claude Code added fallback models, cross-session message security, and broader deny-rule globs.

Anthropic's early-June updates bring fallback model support, broader deny-rule glob matching, stronger cross-session message security, and more reliable thinking controls. The cross-session hardening and deny-rule globs specifically target safer unattended and multi-agent runs. Incremental, but if you run Claude Code in a pipeline like I do, fallback models alone are worth turning on, a stalled primary model no longer kills the whole run.