Anthropic users now face a critical choice: opt in or out of having their Claude conversations used for AI training by September 28. Previously, user data was automatically deleted within 30 days unless flagged for policy reasons, or stored for up to two years. The new policy allows conversations to be retained for five years if the user does not opt out. Business users of Claude for Work, Claude Gov, and Claude for Education are not affected.

The company claims that allowing user data to be used will improve Claude’s safety, coding, and reasoning abilities. Behind this explanation, the main purpose is to gather large-scale real-world data that can boost Anthropic’s competitiveness against OpenAI and Google. Access to such data ensures the development of higher-quality, more sophisticated AI models.

These policy shifts also highlight a broader industry trend toward stricter data oversight. OpenAI is facing a lawsuit requiring the retention of all ChatGPT conversations, even deleted ones, reflecting ongoing tensions between privacy and AI innovation. Many users remain unaware of these changes and may consent inadvertently, raising concerns about meaningful informed consent.

The company’s interface reinforces this issue, with a prominent “Accept” button and a smaller toggle for data-sharing permissions preset to “On.” Experts warn that the design can lead users to unknowingly grant permission. The policy changes exemplify the ongoing struggle between user privacy, ethical AI practices, and the need for large-scale data collection.