A surprising discovery has shed light on potentially compromising leaks involving OpenAI’s ChatGPT conversations. The leaks, described as ‘cringey chat logs,’ were found in Google Search Console (GSC), a tool predominantly used for monitoring search traffic.
Traditionally, GSC displays search queries based on keywords or short phrases users enter on Google. However, since September, lengthy and personal queries, resembling chat interactions seeking advice on relationships or business matters, appeared in GSC reports. These leaked conversations, likely intended to be private, raised concerns about user privacy and data security.
Analysts from analytics consulting firm Quantable and web optimization consultant Slobodan Manić collaborated to investigate the source of these leaks. Their findings suggested that OpenAI might be scraping Google directly to gather user prompts. Despite OpenAI’s acknowledgment of the issue and subsequent resolution, questions linger regarding the extent of user data exposure and privacy infringement.
This incident underscores the importance of safeguarding user information in AI-driven interactions. Transparency and accountability in data handling practices are crucial to maintaining trust in AI technologies.
Source: Ars Technica