OpenAI has shut down a recently introduced option in ChatGPT that let users make shared conversations visible to search engines, after outside researchers found thousands of supposedly private chats appearing in Google results. The company said the opt-in setting was part of a “short-lived experiment” aimed at helping people discover useful exchanges, but conceded it created “too many opportunities for folks to accidentally share things they didn’t intend to.” To expose a chat, users had to click the Share button, generate a link and tick a box marked “Make this chat discoverable.” Security specialists demonstrated that a simple query such as “site:chatgpt.com/share” surfaced sensitive material ranging from health inquiries to job-application details. An archived scrape on Wayback Machine listed more than 10,000 indexed links. Chief information security officer Dane Stuckey said OpenAI has removed the feature for all accounts and is “working to remove indexed content from the relevant search engines,” a process expected to finish by 1 August. The company did not disclose how many users had opted in or how long the links had been available. The reversal highlights the tension generative-AI companies face in balancing knowledge-sharing tools with privacy safeguards. Google, Meta and other providers have faced similar backlash after experimental AI features exposed user data, prompting calls for stronger default protections and clearer disclosures.