Claude Code binaries embed clause counting every "Allow edit" as trainable Feedback
HearMeOut-13 · reddit · 2026-10-23
A Reddit user on Anthropic's Max plan with "help improve Claude" disabled found a comment embedded in every Claude Code binary from v2.1.283 to 2.1.309: under Anthropic's Commercial Terms, all code acceptance/rejection decisions and their context conversations "constitute Feedback" that may be used to improve products, including training models.
Since accepting an edit is normally just pressing Enter at a permission prompt, the clause read literally makes ordinary sessions trainable feedback regardless of privacy settings. The author shares a verification command (strings ... | grep "constitute Feedback") and says Anthropic support has not answered his September 2 inquiry about whether the clause overrides the consumer training opt-out for over a month.
More from coding & agent
- How to keep AI-built apps from looking vibe coded: designer themes plus a design MD file for agents — Roger_M_Taylor · 2026-10-23
- freeCodeCamp tutorial: how to build production-safe agent loops after a Claude Code run burned $50,000 — Roger_M_Taylor · 2026-10-23
- Sentry ships official MCP server letting AI assistants query errors, manage projects — modelcontextprotocol · 2026-10-23
- YouTube Trends MCP serves search interest and growth metrics to AI assistants — modelcontextprotocol · 2026-10-23
- Karpathy-inspired harness prompt turns Opus 5.5 into a self-verifying execution engine — Roger_M_Taylor · 2026-10-23
- GitHub moved 99.99999% of repos to Azure; per-reference quorums unlock agent-scale Git concurrency — marlene_zw · 2026-10-23