Anthropic says it will train its AI models on user data, including new chat transcripts and coding sessions, unless users opt out by September 28
You can choose to opt out. … Anthropic will start training its AI models on user data, including new chat transcripts and coding sessions, unless users choose to opt out.
Context & Ripple Effects
Anthropic’s change turns its consumer chat and coding products into a larger feedback source for model development, while placing the choice inside a revised-terms flow. The closely related coverage says users must accept those terms and make the choice by the September 28 terms deadline.
The move contrasts with an earlier enterprise-facing boundary: OpenAI said API submissions would not be used for model training unless customers opted in, following criticism. Anthropic’s policy therefore sharpens the distinction between protections users may expect in paid or developer workflows and defaults applied to product interactions.
First-order effects
- Anthropic users’ new chats and coding sessions become eligible for model training unless they opt out by September 28, making privacy settings an immediate condition of continued use.
- Anthropic gains a potential stream of real-world conversational and coding interaction data; users who do not want that use must actively register their preference.
Second-order effects
- AI coding and assistant providers face more pressure to make their training defaults, retention practices, and opt-out paths legible, especially for users handling sensitive work in chat interfaces.
- The policy raises the value of products and tiers that can credibly separate customer inputs from model-improvement data, echoing OpenAI’s earlier API opt-in position.
Third-order effects
- If opt-out-by-default training becomes common, access to behavioral and workflow data could become a more important competitive advantage for AI assistants, alongside model quality and distribution.
- The recurring consent question may push the market toward clearer separation between consumer, workplace, and developer data policies; the exact balance will depend on user opt-out behavior and privacy expectations.
The trend: This is one data point in the commercialization of AI interaction data, as assistants seek to turn everyday chats and workflow activity into model-improvement inputs under configurable consent rules.