ANALYTICAL MEMORANDUM on the Capabilities of Large Language Models (LLMs) for Inferential Psychometric Profiling of Users and the Resulting Systemic Risks for Society and the State by Peter Eidos.

Dear Sir or Madam,I am submitting this memorandum to draw attention to a critical area in the development and application of Artificial Intelligence that—despite its importance— remains largely outside public debate. This concerns the ability of large language models (LLMs) to infer (guess with exponentially increasing accuracy) psychological traits, cognitive styles, motivational structures and potential patterns of thought of users, even with minimal input data. In recent months, there has been a rapid advancement in the inferential capabilities of these systems, enabling them not only to answer questions but also to deduce deep user properties from the way language is used. This phenomenon requires urgent regulatory reflection, as current legal frameworks do not explicitly cover this category of information processing. These capabilities have both positive implications (e.g., in licensed diagnostics, therapeutic support or adaptive education) and highly risky ones, particularly in the context of public institutions, democratic processes, and surveillance systems.