OpenAI Disables ChatGPT Share Links After 100,000 Chats Indexed by Google

OpenAI has removed ChatGPT’s public “share” link function after researchers found almost 100,000 user conversations had been indexed by Google search results, exposing everything from draft business contracts to personal relationship disclosures. The links, which users created voluntarily to show chats to third parties, were set to be publicly reachable; Google’s crawler treated them like any other web page. The cache was first compiled last week by a security researcher who used standard search queries to collect the URLs and assemble a searchable archive. Privacy specialists said the material contained commercially sensitive data as well as intimate personal information, illustrating how generative-AI tools can become inadvertent repositories of confidential content when default-sharing settings are misunderstood. OpenAI confirmed the exposure and, according to chief information security officer Dane Stuckey, has now disabled the sharing option while it works with search engines to purge the links from their indexes. The company is also urging users who previously shared chats to review and, if necessary, delete the published URLs. The incident adds to mounting scrutiny of generative-AI platforms over data security. Regulators and corporate customers are likely to press OpenAI and rival providers to adopt stricter privacy-by-default designs as the technology’s user base rapidly expands.