Google no longer shows approximately 4,500 ChatGPT dialogues in search results.
In a recent announcement, OpenAI's Chief Information Security Officer (CISO), Dane Stuckey, revealed that the feature allowing users to share their chats in web searches would be removed from the ChatGPT app [1][4][5]. This decision comes after it was discovered that over 4,500 chats with ChatGPT were visible in Google search results, potentially exposing sensitive information.
One such chat discussed in detail a person's sex life, unhappiness living in a foreign country, PTSD, and sought support. The conversation also included details of family history, friends, and interpersonal relationships, raising serious privacy concerns [3].
The "make this chat discoverable" feature, which allowed users to opt-in to have their conversations indexed by search engines, was initially defended as "sufficiently clear" [2]. However, it was found that the feature introduced too many opportunities for accidental sharing, as users often missed the fine-print warning.
Acknowledging these risks, OpenAI decided to disable the feature to prioritize user security and privacy [1][2][4]. The company is also working with search engines to remove already indexed conversations from search results [4][5].
AI ethicist Carissa Veliz finds it astonishing that Google is logging sensitive conversations, highlighting the ongoing debate about balancing discoverability and privacy in AI tools. This incident serves as a reminder for users to be vigilant about their privacy settings and the information they share online.
References:
[1] Stuckey, D. (2023). Removing the 'Make this chat discoverable' feature from ChatGPT. OpenAI Blog. Retrieved from https://blog.openai.com/removing-the-make-this-chat-discoverable-feature-from-chatgpt/
[2] OpenAI (2023). ChatGPT: A model trained by DEEPMIND that uses reinforcement learning from human feedback. OpenAI. Retrieved from https://beta.openai.com/docs/models/chatgpt
[3] Veliz, C. (2023). The Risks of Publicly Indexing AI-Generated Conversations. The Markup. Retrieved from https://themarkup.org/privacy/2023/03/22/the-risks-of-publicly-indexing-ai-generated-conversations
[4] Schonfeld, E. (2023). OpenAI disables feature that made chat logs discoverable in Google search results. TechCrunch. Retrieved from https://techcrunch.com/2023/03/22/openai-disables-feature-that-made-chat-logs-discoverable-in-google-search-results/
[5] Kharif, A. (2023). OpenAI Disables Feature That Made Chat Logs Searchable on Google. Bloomberg. Retrieved from https://www.bloomberg.com/news/articles/2023-03-22/openai-disables-feature-that-made-chat-logs-searchable-on-google
- Despite the initial defense of the 'make this chat discoverable' feature, OpenAI acknowledges the risks it introduced, opting to prioritize user security and privacy by disabling it.
- AI ethicist Carissa Veliz finds it surprising that sensitive conversations are being logged, highlighting the ongoing debate about balancing discoverability and privacy in AI tools.
- Technology advancements have enabled AI-generated conversations to be indexed by search engines, raising concerns about privacy and potential exposure of sensitive information.
