ChatGPT and Microsoft record our private life | Press review n°1
A selection of important news about artificial intelligence during the week of April 7 to 13, 2025.
Welcome to the first press review of Artificial reality. I will publish a selection of the latest important developments in AI every week. Have a good read!
📰 Read
ChatGPT now records all conversations
OpenAI CEO Sam Altman announced Thursday on X that ChatGPT can now reference all the conversations of the users:
we have greatly improved memory in chatgpt--it can now reference all your past conversations!
this is a surprisingly great feature imo, and it points at something we are excited about: ai systems that get to know you over your life, and become extremely useful and personalized.
The purpose of this feature is to improve the chatbot’s personalization, which “gets to know” the user through conversations thanks to a long-term memory in order to give answers tailored to his interests and preferences.
This is an update from the “Memory” option that was added to ChatGPT last year. It allows users to ask the chatbot to memorize certain questions, prompts and settings for subsequent replies.
With this new functionality, ChatGPT will now remember all conversations, even if the user did not choose to record them. It is however possible to disable this option in the settings.
This long-term memory is already available for Pro subscribers ($200 a month) and will soon be rolled out for Plus subscribers ($20 a month). There are exceptions: the update will not be available for the moment in the European Union, the UK, Switzerland, Norway, Iceland and Liechtenstein, probably because these regions have tight AI regulations, according to The Verge.
The new feature announced by Sam Altman intensifies the mass surveillance put in place by technology compagnies at the dawn of the digital age. The recording of all conversations with this chatbot, including our most personal questions, does represent a further loss of privacy.
This personalization of ChatGPT at all costs could also be a source of vulnerabilities. By getting to know us through our conversations, the chatbot will also learn our biases, our prejudices, our desires, our weaknesses. It could then influence our decisions and even manipulate us. Because technology is never neutral, it reflects and strengthens the ideology of its creators and directors.
Read this tweet by Edward Snowden.
Microsoft rolls out an AI program seen as a “potential privacy nightmare”
A controversial AI-powered tool is currently being rolled out by Microsoft. Called Recall, this program takes automatic screenshots every few secondes on Copilot+ PCs and laptops, a line of computers from the American company.
The purpose of Recall is to allow clients of Microsoft to search the history of their computer. The corporation gives the example of a person who saw a dress on an online store a few days earlier and who could then easily find it again with a Recall search.
The announcement of this feature raised a lot of concerns last year, an AI and privacy expert labelling it as a “potential privacy nightmare” because Recall can record very sensitive informations such as private emails, messages, photos and videos, passwords, credit card number, tax reports, etc. Microsoft then decided to postpone its launch for a few months.
A preview mode of this feature is now available to some people. The company has since decided to make it an opt-in option (users have to choose to activate it), while it was initially planned to be turned on by default for everyone. The tool will be rolled out worldwide, including in the European Union later this year.
Microsoft defends itself against the criticism regarding the intrusion into the private life of its clients, arguing that the screenshots are stored locally on the computer of the users and that it will be necessary to verifier one’s identify before being able to access the images.
If Microsoft says the truth, it remains possible that these screenshort will be stored on one of its servers later on, thus becoming accessible to the corporation and to the other entities (companies, government agencies, etc.) with which it shares data.
In any case, this tool represents a loss of privacy because personal and sensitive informations will be recorded, often without the consent of all the people involved. Let’s take the example of a person using Recall who makes automatic screenshots of the messages she exchanges with a friend, without the latter having given her permission that this conversation would be memorized by a program.
These screenshots could also be stolen in cyberattacks. They thus create new vulnerabilites for users.
DOGE uses AI to surveil federal workers
The Departement Of Government Efficiency, led by businessman Elon Musk, is using an artificial intelligence system to surveil the employees of a federal agency, Reuters revealed on Tuesday.
At the Environmental Protection Agency, managers were informed that Elon Musk’s team was deploying an artificial intelligence to monitor if some workers criticized Donald Trump or Elon Musk in messages or during videocalls.
Reuters calls this practice “an extraordinary use of technology to identify expressions of perceived disloyalty”. Cybersecurity experts and government ethicists think that the Trump administration could use the informations gathered with this AI to go after political targets.
🎥 Watch
Democracy Now! published a video on Friday about two Microsoft employees who protested against the use of the company’s artificial intelligence and cloud by the Israeli army during the war in Gaza.
“We wanted everyone to know that Microsoft’s cloud and AI are the bombs and bullets of the 21st century”, explained Vaniya Agrawal, a former Microsoft employee who was fired after disrupting an event of the corporation on April 4.
Thank you for reading the very first press review of Artificial reality! Have a good week and see you next Sunday.