PSA: Your Claude shared chats and Artifacts may have ended up on Google
Shared Claude AI chats and Artifacts were recently found to be publicly searchable on Google, potentially exposing sensitive user data including health records and private company documents. While Anthropic, Claude’s developer, suggested user error for links posted publicly, Google noted it respects site owners’ indexing directives.
Numerous shared Claude AI chats and Artifacts – interactive mini-apps and documents – were inadvertently made public on Google. This exposure, which included potentially sensitive information such as health records, private company documents, and personal details, highlights concerns surrounding data privacy in AI interactions. The issue appears to stem from Claude’s "share chat" feature, which generates public links for conversations.
Anthropic, Claude’s developer, attributed the visibility of these shared links in search results to users posting them on public forums or social media. They emphasized that privately shared links remain unsearchable. Google, in response, clarified that it indexes pages based on site owners' directives and does not control public web content, stating it respects indexing controls set by website administrators.
The problem was initially identified by a Reddit user and subsequently reported by 404 Media before TechCrunch’s investigation. Although the issue seems to have been largely remediated, previous reports indicated the exposure of highly sensitive information. This incident echoes a similar situation last year where hundreds of Claude chats were indexed by search engines, underscoring ongoing challenges in managing privacy with AI-powered tools.
Related articles
Microsoft launches its first cybersecurity model, plus a new agentic cybersecurity system
Microsoft has unveiled MAI-Cyber-1-Flash, its first cybersecurity-specialized AI model, designed to uncover vulnerabilities in complex codebases. Complementing this is Perception, a new platform that deploys AI agents to automate security workflows and mitigate bugs. Microsoft claims these tools offer superior performance and cost-efficiency compared to competitors, aiming to combat AI-powered cyber threats with advanced AI defenses.
Google’s deepfake detector system used to debunk McConnell hoax pic
Google's SynthID, an invisible watermarking system, successfully debunked a widely shared AI-generated image of Senator Mitch McConnell. This marks a significant victory for anti-deepfake technology, proving SynthID's effectiveness in identifying AI-generated content across various platforms.
Helping K–12 educators build practical AI skills
OpenAI Academy and the Walton Family Foundation are launching an "AI Skills Jam" for K–12 educators this summer. This initiative aims to equip over 1,600 teachers and administrators with practical AI skills to enhance teaching, planning, and communication, addressing the challenge of rapid technological change in education.