Generative AI become personalized cognitive partners thanks to user memory. But this personalization making any change in model very difficult.
Generative artificial intelligence has already infiltrated at a dazzling speed in all the spheres of our existence. From students to professionals, from creators to daily users, the use of LLM is generalized, transforming the modes of reasoning, writing, research, learning and interaction with information.
This massive adoption, accelerated by tools like Chatgpt or Grok, marks a turning point where AI is no longer a punctual auxiliary, but a digital interlocutor integrated into our personal, intellectual and social routines.
AI user memory to become your arm
At the center of this integration is increasing personalization, based on “user memories”, the data set accumulated over the interactions, including historical conversations, preferences and specific contexts. By relying on this memory, AI is capable of considerably improving the quality, precision and relevance of its responses. She can adapt her style, refine her suggestions, avoid repetitions and take into account implicit preferences, thus offering a more fluid, consistent and efficient experience. This adaptability creates an environment of tailor-made use. But at the cost of a strong informational dependence.
Cognitive and relational dependence
Because this personalization creates an implicit, but deeply anchored link, between the user and his AI assistant. The tool is no longer content to respond to isolated requests; He anticipates, adapts and enriches exchanges by relying on an intimate knowledge of the individual. What might seem an ergonomic advance in reality conceals a relational and cognitive dependence, where each user invests time and data to “train” their AI. Imagine an exceptional right arm that you have formed for 2 years and which, for one reason or another, can no longer work with you. You will have to reform a right arm. The transition from one AI to another is expensive in time and loss of efficiency during the transition.
Already all veroused?
Since when do you train your AI? Have you ever noticed how it improves? What if you wanted to change it? Without interoperability of user profiles, we are witnessing the emergence of a structural, invisible but deeply massive lockout. This locking is not only technical; It is relational, by hampering the fluidity of increased human interactions; cognitive, by capturing the thinking and work patterns in a closed ecosystem; And personal, by linking the intellectual and expressive habits of an individual to a single supplier.
Ultimately, this could consolidate digital monopolies, where great actors-often extra-European-dictate the rules, to the detriment of competition and individual freedom.
Europe already has the tools
Europe, strong of a legal and economic tradition anchored in interoperability and portability, has levers to counter this drift. Recall that, as early as 2003, the portability of mobile phone numbers released consumers from operators’ channels, promoting real mobility, innovation and lower prices. More recently, the DSP2 directive has opened bank data via Open Banking, stimulating financial innovation while protecting users. In the health field, initiatives such as my health space in France and the European area of health data (EHDS) provide a secure circulation of sensitive information, placing the citizen at the center of the system.
Finally, the regulation on data governance (DGA) and the future regulation on European data space already impose the interoperability of digital systems, erecting the free movement of data in principle cardinal … as long as the property is respected. This tradition, which aims to prevent opaque dependencies towards technological giants, must now extend to user profiles of generative AI, our piece of outsourced brain.
It is necessary to act quickly and with determination. In fact, yes, AI and its uses are already very impressive, but we are actually only at the beginning.
Europe must impose regulation requiring interoperability and portability of user profiles. Not so much for sovereignty, an important concept if it is seen in the light of economic power and not of regulation, but rather to guarantee freedom for citizens to have the piece of brain, theirs, which they inject in an LLM or another. Such a measure not only would preserve market balance, but would also strengthen confidence in these technologies, aligning them with the founding principles of the European Union.
