Most AI chatbots devour your user data – these are the worst offenders

Like many people today, you may turn to AI to answer questions, generate content, and gather information. But as they say, there’s always a price to pay. In the case of AI, that means user data. In a new report, VPN and security service Surfshark analyzed what types of data various AIs collect from you and which ones scoop up the greatest amount.

For its report, Surfshark looked at 10 popular AI chatbots — ChatGPT, Claude AI, DeepSeek, Google Gemini, Grok, Jasper, Meta AI, Microsoft Copilot, Perplexity, Pi, and Poe. The analysis checked the privacy details for each app on Apple’s App Store as well as the privacy policies for DeepSeek and ChatGPT. The goal was to determine how many types of data each app collects, whether it gathers data linked to you, and whether the app uses third-party advertising.

Also: The best AI chatbots: ChatGPT, Copilot, and notable alternatives

Surfshark focused on 35 different data types, including contact info, health and fitness, financial info, location, sensitive info, contacts, user content, history, identifiers, diagnostics, usage data, and purchases. As one example, sensitive info includes racial or ethnic data, sexual orientation, pregnancy or childbirth information, disability, religious or philosophical beliefs, trade union membership, political opinion, genetic information, or biometric data.

All 10 AI apps collect some type of user data. The average number of data types amassed was 13 out of 35. Some 45% of the apps gather your location. Almost 30% track user data, which means that the information collected from the app is linked with third-party data to deliver targeted ads or share with a data broker.

So, who are the worst offenders?

According to Surfshark’s examination, Meta AI came in first place. This one collects the most user data, scooping up 32 out of the 35 possible types, or 90% of them. Meta AI was the only AI app that snagged data across categories such as financial information, health and fitness, and sensitive information. Further, only Meta and Copilot grabbed data linked to the user’s identity to display third-party ads. Meta AI can also gather as many as 24 different data types for this purpose.

Next on the list was Google Gemini, which collects 22 different data types. The app grabs data on your precise location as well as contact info (name, email address, phone number, etc.), user content, contacts (a list of contacts on your phone), search history, and browsing history.

Poe, Claude, and Copilot rounded up the top five. Poe collects 14 different types of data, Claude 13, and Copilot 12. Poe and Copilot are two out of three apps (Jasper being the third) that gather data used to track you. By using your device ID, Poe and Copilot can sell your data to brokers or use it to display targeted ads in the app.

In sixth place was DeepSeek, a Chinese AI that has garnered praise for its performance but concerns over its roots. In Surfshark’s analysis, DeepSeek wound up in the middle, collecting 11 types of data, including your chat history.

However, the problem with DeepSeek is two-fold. First, since it originates in China, your results could be censored with blocks on restricted content. Second, the data collected by the AI is sent to China Mobile, a Chinese government-operated telecom company banned from operating in the US since 2019 over national security concerns.

Also: How Apple plans to train its AI on your data without sacrificing your privacy

Surfshark also notes that DeepSeek claims to hang on to information as long as needed, saving it on servers located in the People’s Republic of China.

“Don’t let your guard down, as chats stored on servers are always at risk of being breached,” Surfshark cautioned in its report. “According to The Hacker News, DeepSeek has already experienced a breach where more than 1 million records of chat history, API keys, and other information were leaked. It is generally a good idea to be mindful of the information provided.”

Next up is ChatGPT, gathering 10 types of data, such as contact information, user content, identifiers, usage data, and diagnostics. On the positive side, OpenAI’s AI doesn’t track your data or use third-party ads. Plus, privacy-minded users can always turn to temporary chats in which all data is automatically deleted after 30 days. Even further, you can request that your personal data not be used for training purposes.

Completing the top 10 list were Grok, Pi, and Jasper. Surfshark found that Grok gathers seven types of data, while Pi and Jasper each scoop up five different types. Though it came in last place, Jasper collects device IDs, product interaction data, advertising data, and other usage data to send you targeted ads or share data with a broker.

Also: Want AI to work for your business? Then privacy needs to come first

Collecting user data is a common practice not just among AI bots but among other mobile apps, social media sites, search engines, and software. Often, that’s the price you end up paying to get free or inexpensive products that rely on advertising. But that doesn’t mean you have to submit willingly.

ChatGPT and other AI apps and services do provide ways for you to prevent or at least limit the collection of your data. Your best bet is to investigate the privacy policies and settings for any AI you use to see how you can take charge of your own data.

Want more stories about AI? Sign up for Innovation, our weekly newsletter.