Over the weekend, Redditors discovered that Claude AI chats were accessible via Google Search. While Anthropic quickly patched the issue, it demonstrates the privacy risks of AI programs.

Researchers reveal Anthropic's Claude Co-Work exploited a Linux flaw to escape its sandbox, echoing recent security concerns surrounding OpenAI's ChatGPT agent vulnerabilities.

Shared conversations with Anthropic's Claude chatbot briefly appeared in Google search results because the pages lacked a noindex tag. Users said some chats contained crypto keys…