Particle.news

Anthropic Sets Sept. 28 Default to Train on Most Consumer Claude Chats

The shift makes training the default for consumer chats, raising questions about meaningful consent.

Overview

  • The change covers Claude Free, Pro, Max, and Claude Code users, with enterprise, government, education, and API access via Amazon Bedrock or Google Vertex excluded.
  • Existing users will be included unless they opt out by September 28, and new users can disable the “Help improve Claude” toggle at sign-up or later in privacy settings.
  • Anthropic will retain data from opted-in accounts for up to five years, replacing the prior 30‑day retention policy.
  • Only new and resumed conversations from opted-in users will be used for training, leaving older chats out of scope.
  • Anthropic cites safety and capability gains from real‑world data, as critics fault the opt‑out design and experts note FTC warnings about unclear policy changes.