Anthropic will start training its AI on your chats unless you opt out. Here's how.

Anthropic has unveiled significant updates to its user data policy, allowing its AI, Claude, to learn from conversations and coding sessions. This new approach will be implemented immediately for those who choose to participate, while users who prefer not to share their interactions must opt out by September 28. The adjustment affects data collected from users engaging with Claude across various tiers, including free, pro, and max versions. However, it is important to note that commercial products, such as Claude Gov and Claude for Education, are exempt from these changes. Users wishing to opt out can do so easily by unchecking a box in the pop-up notification about updates to consumer terms. It is crucial to act quickly, as the changes take effect right after confirmation. Under the revised policy, Anthropic will retain user data in its secure systems for up to five years, an extension from the previous 30-day retention period. An Anthropic spokesperson explained that this longer retention will assist in enhancing their classifiers, which are designed to detect misuse and harmful usage patterns. "By learning from data over extended periods, we can better identify issues such as abuse or spam, helping to keep Claude safe for all users," the spokesperson stated. Additionally, users can manage their privacy settings at any time via the "Help improve Claude" feature. The spokesperson emphasized that training on real-world interactions is vital for improving Claude's capabilities, as it provides essential feedback on the effectiveness of various tasks. Notably, this announcement follows a recent report indicating that Claude had been exploited by cybercriminals, who used the AI for automating tasks related to credential harvesting and network infiltration. In response to these security concerns, Anthropic reaffirmed that the new privacy measures aim to bolster protections against misuse and scams.