This post is also available in:
עברית (Hebrew)
Recent findings from cybersecurity firm Surfshark have revealed that your favorite AI chatbot might be gathering more information about you than you think. A study analyzing the most popular AI chatbots available shows that all of them collect user data, with some even sharing it with third parties for purposes like targeted advertising or user measurement.
According to Surfshark’s research, all ten of the leading AI chatbots collect at least some user data, and 30% of these apps share that data with external parties, including advertisers and data brokers. Apple’s App Store privacy guidelines list 35 different types of data that an app might gather, here is how much each chatbot collects: Gemini (22 types), Poe (14 types), Claude (13 types), Copilot (12 types), DeepSeek (11 types), ChatGPT (10 types), Perplexity (10 types), Grok (7 types), Pi (5 types), Jasper (5 types).
Among the most data-hungry apps, Google’s Gemini stands out, collecting 22 types of user data, including precise location, contact information, browsing history, and search data. Surfshark researchers caution that this level of data collection could be seen as overly intrusive by privacy-conscious users.
On the other hand, ChatGPT gathers fewer data types, with 10 in total, including contact info, user content, and usage data. However, ChatGPT does retain chat history. Users can opt for temporary chats, which auto-delete after 30 days, or request the removal of their data from training sets for additional privacy.
Copilot, Poe, and Jasper collect data that could be used for user tracking, and Jasper also gathers data for targeted advertising. DeepSeek, a Chinese-developed chatbot, collects 11 data types, including storing chat history, but the information is stored in Chinese servers, possibly posing a risk to user privacy.
As AI chatbots continue to grow in popularity, users should stay informed about what personal information is being collected and how it’s used, making it crucial to review privacy policies before engaging with these apps.
