Private chats from Anthropic's Claude AI appeared in Google search results
Users' confidential conversations surfaced publicly online
1Scope of the leak
Private conversations with Anthropic's Claude AI chatbot unexpectedly appeared in Google search results. The exposed data included cryptocurrency wallet keys, personal details, and even erotic content. Only chats using the 'share' feature were affected.
2How it happened
Links generated through Claude's share button were indexed by Google. A specific search query shared on Reddit revealed hundreds of shared conversations. While links weren't guessable, once shared they became publicly accessible.
3Company response
Anthropic clarified they don't automatically share chat directories with search engines. Emphasizing that users must consciously share links, the company claims to have fixed the issue. However, some previously shared links remain active.
4Industry-wide issue
This isn't the first occurrence: OpenAI's ChatGPT and Elon Musk's Grok experienced similar leaks. AI chat share URLs can inadvertently become exposed to search engines. The industry has yet to find a permanent solution.
5Privacy implications
Users often discuss health, legal, and work matters with chatbots. Shared conversations can be more sensitive than regular documents. Such leaks risk exposing vast amounts of personal data unexpectedly.
6User precautions
Experts advise caution when sharing sensitive information via chatbots. Share links should be distributed selectively and revoked when unnecessary. Companies also need to redesign sharing mechanisms to prevent accidental exposure.
Original article: Fortune | FORTUNE — Anthropic'in Claude sohbetleri Google'da sızdı