Meta AI and ChatGPT: Champions of Personal Data Collection
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisis ton rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
A Massive and Widespread Data Collection Among AI Chatbots
A recent study conducted by Surfshark highlights the extent of data collection by major artificial intelligence chatbots. The analysis, based on App Store privacy information and service policies, reveals that all the applications studied collect at least one type of user information. On average, 14 types of data are harvested from the 35 identified categories.
Among the most frequently collected data, about 70% of applications retrieve users' location. Other information, such as contact details, generated content, as well as search or browsing history, is also commonly exploited. This data is not only used for the operation of services but is also utilized for analysis, personalization, and marketing purposes, increasing the risks in the event of a data breach.
Surfshark warns of potential dangers: “Do not let your guard down: conversations stored on servers remain exposed to the risk of leakage.” This alert echoes an incident involving DeepSeek, which suffered a data breach that resulted in the disclosure of over a million records, including chat histories and API keys.
Contrasting Practices Among Players
Surfshark's study also reveals significant differences in the practices of various chatbot services. Meta AI stands out as the most intrusive service, collecting 33 types of data out of 35 possible, nearly 95% of the categories. It is the only one to include financial data and, along with Gemini, to collect sensitive information such as political opinions, biometric data, and sexual orientation.
In contrast, ChatGPT has expanded its data collection, increasing from 10 to 17 types of information. Among these data are approximate location, audio data, search history, and certain advertising-related information. Conversely, Claude maintains a more stable collection with 13 types of data, primarily necessary for the operation of the service.
The ranking of AIs that collect the most data, as well as the nature of the data harvested, is detailed in Surfshark's study.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.