Хочете будувати кар’єру в Data Science та AI? Vodafone Big Data Lab відкриває новий набір —>
Оксана СтепураAI Eng
26 October 2026, 16:50
2026-10-26
ChatGPT creates psychological portraits of users without their request. What did the researchers find out?
Researchers found information about users' desires, emotions, intentions, and beliefs in 52% of ChatGPT's memory records. 96% of these records were created by the chatbot itself, without asking users to remember anything.
Researchers found information about users' desires, emotions, intentions, and beliefs in 52% of ChatGPT's memory records. 96% of these records were created by the chatbot itself, without asking users to remember anything.
This is reported by Futurism, citing a study by scientists published in the proceedings of the ACM Web Conference 2026.
To find out what ChatGPT knows about its interlocutors, the researchers studied the chat histories of 80 users. In the conversations of 65 of them, they found 2,050 memory entries — a kind of notes that the AI keeps to better understand the person in subsequent conversations.
The AI assistant's notes also include quite personal details. Examples cited by the researchers include a person feeling lonely despite having friends and family, fearing failure when making important career decisions, or relying on the chatbot's advice due to limited access to doctors.
In addition, 28% of the records contained personal data as defined by the European GDPR regulation. This could include names, addresses, information about a person's employment or social status. Another 7% of the records contained particularly sensitive information, such as information about health, religious or political beliefs.
ChatGPT most often retained users' desires - they appeared in 73% of entries with psychological information. In contrast, emotions appeared in only 14%. Scientists suggest that the chatbot prioritizes a person's long-term goals and intentions, rather than their fleeting experiences.
Sometimes ChatGPT even remembered things that users didn't actually say. For example, one person asked for recommendations for bands similar to Nirvana. The chatbot decided that she liked Nirvana and stored that in its memory. Even though the user didn't directly state anything like that.
Previously, dev.ua wrote that OpenAI introduced an updated ChatGPT memory system called Dreaming V3. It independently analyzes the communication history, remembers important details about the user, and can update information that has lost relevance.
«Чи є у мене талант, якщо комп’ютер може імітувати мене?». Штучний інтелект пише книги авторам Amazon Kindle. The Verge поспілкувався з авторами та виявив багато цікавого
Письменники-романісти використовують штучний інтелект для створення своїх творів. Видання про технології The Verge поспілкувалося з письменницею Дженніфер Лепп, яка випускає нову книгу кожні дев’ять тижнів, й дізналося про те, як працює штучний інтелект для написання романів. Наводимо адаптований переклад статті.
Have important news to share? Message our Telegram bot
Key events and useful links in our Telegram channel