Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
People who use AI chatbots often disclose more data than expected. VPN provider Surfshark took a close look at the privacy practices of the ten most popular chatbot apps in the Apple App Store – with some alarming results. ChatGPT in particular stands out: The app collects significantly more data today than it did a year ago. And Meta AI is the undisputed leader.
For the analysis, Surfshark evaluated the data protection details of the ten most-used AI chatbot apps in the Apple App Store. Apple defines 35 possible data types that apps can collect – from location data and health information to financial data. On average, the apps analyzed collect 14 of the 35 possible data categories.
Important to note: Not all data collection is problematic per se. Many types are simply necessary for the operation of the app – for example for user authentication, fraud prevention or customer support. The decisive factor is which data is also used for advertising, tracking or disclosure to third parties.
Unsurprisingly, but still remarkable: Meta AI collects by far the most user data of all the apps analyzed, with 33 out of 35 possible data types. This corresponds to almost 95 percent of all possible data categories. Meta AI is also the only app that grants itself the right to analyze users’ financial information.
Together with Google Gemini, Meta AI also accesses particularly sensitive data – including ethnic origin, sexual orientation, political views, genetic information and biometric data. Anyone using Meta AI should be aware of this far-reaching data processing.
Google Gemini is in second place with 23 out of 35 data types. What is particularly striking here is the combination of precise location data, search history, browser history and contact data – a depth of data that reflects Gemini’s proximity to Google Search and the Chrome browser. Only four of the ten apps examined collect precise location data: Gemini, Meta AI, Copilot and Perplexity.
The development at ChatGPT is particularly striking: OpenAI’s chatbot has increased the number of data categories recorded by 70 percent compared to the previous year – from 10 to 17 types. New additions include rough location data, health and fitness data, search histories, audio data and advertising data.
This fits in with OpenAI’s latest strategic decisions: The group is working on its own browser, is increasingly pushing into the health segment and is planning advertising revenue. All of this is now also reflected in the privacy policy. According to Apple, 14 of the 17 types of data collected are intended for app functionality – but also for analytics, personalized advertising and third-party advertising.
Anthropic’s Claude performs comparatively well with 13 out of 35 data types. According to Surfshark, the data collected is primarily required for app functionality. However, this data can also be used for analytics and the developer’s marketing – including rough location data and uploaded content such as photos or videos.
DeepSeek also collects 13 types of data, including rough location data and search histories. The key difference: as a Chinese company, DeepSeek is not subject to European data protection laws such as the GDPR. The data is stored on servers in China. In addition, DeepSeek was once involved in a data scandal in which over one million data records with chat histories, API keys and other information were publicly accessible.
You can find more background information on DeepSeek in our article DeepSeek: China’s open source AI as a challenge for ChatGPT and co.
Surfshark’s study makes it clear that AI chatbots are no longer simple question-and-answer tools, but platforms that penetrate deep into users’ everyday lives – and want to collect a lot of data accordingly. The increase in ChatGPT in particular shows where the journey is heading: the broader the offering, the more data is collected.
Anyone who values data protection with AI chatbots should at least take a look at the privacy policy – and think carefully about what information (documents, health data, personal details) they actually enter into the chat. Surfshark has published the complete data table in a public Google spreadsheet.
Source: Surfshark Research
Your email address will not be published. Required fields are marked *
Contact us via one of our social media channels to send us your comments or suggestions for improvement. We look forward to your message!
You are currently viewing a placeholder content from Facebook. To access the actual content, click the button below. Please note that doing so will share data with third-party providers.
