independent · not affiliated with Anthropic · nothing leaves your browser

Claude Code keeps blocking you with a "Usage Policy violation" — can you recover, or do you have to escalate?

In the June 2026 wave (≈ 2.1.1612.1.167) this false-positive poisons the whole session: once it fires, even "hi" gets blocked. The single thing that decides whether you can fix it yourself is where the state lives. Three questions and one test sort that out.

First, the trap to avoid: this is not driven by your vocabulary. In the June wave it fires on ordinary CRUD edits, a bare git worktree status, even "hi" — turns with zero security content. So rephrasing or sanitizing your prompts will not help, and "what did I say wrong" is the wrong question. The right question is the one below.

This page describes user reports and operator-side steps; it is not account or legal advice. For Usage Policy questions, confirm against Anthropic's own documentation and support. Reproductions referenced: anthropics/claude-code #65866, #65846, #65867, #65873.