From tool to partner

Over recent decades, AI’s rapid development has been matched by shifts in its specialized vocabulary. The terminological unit AI Agent has become central to how systems are described, built, and regulated, yet research is split. Technical histories track capabilities but neglect linguistic construction, while terminological studies are often qualitative and synchronic. This study integrates Frame-Based Terminology (FBT) with a custom NLP pipeline to reconstruct the semantic frame of the “AI Agent” from 1970 to 2026. Using a diachronic Springer Nature corpus, we extract recurrent predicates and participant roles across four phases to trace framing change. Results show ontological drift from a bounded, rule-based tool to a socially situated partner linked to autonomy, alignment, and collaboration. The frame also expands as normative roles become constitutive, highlighting safety and governance concerns and the ethical risks of partner-oriented framings in human–machine ecosystems.