Claude AI will train on your personal conversations by default

Anthropic, the developer behind the widely-used AI chatbot Claude, has announced significant changes to its user data handling practices. Starting September 28, the company will begin training its AI on conversations with users unless they actively choose to opt out. This change marks a departure from Anthropic's previous policy, where user messages were not used for training purposes. Additionally, Anthropic revealed that it will extend its data retention period for messages exchanged with Claude to five years for users who do not opt out of the new training protocols. Previously, the company had a policy of deleting user prompts and outputs within 30 days unless legally required to retain them. However, they did keep user data for up to two years in cases of policy violations. These updates will affect all consumer-tier users of Claude, including those on free and paid plans, when using Claude Code linked to their accounts. However, users on commercial plans such as Claude for Work, Claude Gov, and Claude for Education will not be impacted by these changes, regardless of whether they access the service through third-party platforms like Amazon Bedrock and Google Cloud Vertex AI. Anthropic emphasizes that by allowing their conversations to be used for model training, users will contribute to enhancing the AI's safety features and improving its ability to discern harmful content from benign interactions. The company assures users that this collaboration will lead to advancements in capabilities such as coding, analysis, and reasoning for future iterations of the Claude models. To inform users of these changes, Anthropic will soon display a pop-up notification titled "Updates to Consumer Terms and Policy." This notification includes a default option labeled "You can now improve Claude," which users may inadvertently accept, allowing their data to be utilized in AI training. Users will have the option to defer their decision until September 28, at which point they must make a choice to continue using Claude. If users opt in, their data will begin to be used for training immediately, although this will only apply to new or resumed conversations. However, if a user revisits prior chats and sends a new message, the entire conversation could be included in the training data. For new users, data training preferences will be requested during the sign-up process. Those unaware of the changes may find themselves automatically included in the data training unless they take steps to opt out. To disable participation in the training, users can follow these steps on the website or app to turn off the "Help improve Claude" toggle in their privacy settings.