November 20, 2026
DunavNET’s Srdjan Krčo and Nataša Krčo are among the co-authors of a new position paper published by the AIOTI Working Group Research in November 2026: “GenAIoT – Decentralised Intelligence Systems: The Implications of Generative AI for IoT and Edge Computing.” The paper brings together 27 experts from 13 European countries, under the editorship of Ovidiu Vermesan (SINTEF) and Valerio Frascolla (Intel).
The paper traces the evolution of IoT through five phases: from simple data collection (Connect), through local processing (Compute), task-specific machine learning (Reason) and discriminative AI, to today’s phase of Generative Reasoning. In this new phase, connected systems can interpret context, explain what is happening and propose actions, rather than simply reporting measurements.
To make this possible, the authors argue for a shift away from large, centralised training pipelines towards efficient, localised inference. Hybrid architectures across the IoT–Edge–Cloud continuum can bring AI closer to where data is generated, with potential improvements in latency, privacy, security, resilience and cost.
When AI outputs can influence physical equipment, errors have real consequences. The paper examines risks such as model hallucinations, adversarial attacks and the safety implications of autonomous generative control. A recurring recommendation is to separate probabilistic AI reasoning from physical actuation: the generative model proposes actions, while deterministic safety layers verify them before anything reaches a motor, valve or actuator.
The paper also addresses privacy, data governance, the role of data spaces, alignment with European regulation such as the AI Act, and the standardisation work needed to deploy GenAIoT systems at scale.
The full paper is available on the AIOTI website