A security lapse at Anthropic allowed private Claude AI chat logs to be indexed by Google, making them publicly searchable. The exposed data included conversations that contained cryptocurrency wallet private keys, according to reports that surfaced this week.
What was exposed
The leaked chats were from users of Claude, Anthropic's large language model. Some of the conversations included sensitive information such as crypto wallet private keys, which if still valid could allow unauthorized access to funds. It's not yet clear how many users were affected or how long the data was publicly accessible.
How it happened
The chats were apparently not properly blocked from Google's search crawlers. A misconfiguration in the site's robots.txt or indexing settings likely allowed the conversations to appear in search results. Anthropic has since taken steps to remove the indexed pages and prevent further exposure.
Response and impact
Anthropic acknowledged the incident and said it is working to ensure such leaks don't happen again. The company advised users to rotate any private keys that may have been included in their chats. The timing isn't great for Anthropic, which has been positioning Claude as a secure alternative to other AI assistants.
The incident highlights a broader risk: as more people use AI chatbots for sensitive tasks, the data they share can become a liability if not properly contained. Users are reminded to avoid sharing private keys or passwords in any AI chat platform.
