In this year’s State of the Union speech, the President of the European Commission, Ursula von der Leyen, described artificial intelligence (AI) as a rapidly evolving and fundamental element of Europe’s economy and security. Similar views were also expressed at the 81st United Nations General Assembly. It is now clear that AI is not just a matter of technology, but also of economy, employment, security and governance.
Within this new reality, a particularly significant development is emerging: agentic AI. In Greek, the term agentic AI has been officially rendered as “αυτενεργούσα ΤΝ” (“autonomously acting AI”), a designation originally proposed by the Special Secretariat and later adopted by the European Commission’s translation services. In public discourse, however, the term “AI agents” is also frequently used.
But what exactly are these “agents” in practice? They are systems that go beyond simply answering questions. They understand an objective, plan the individual steps required, and, within defined boundaries, use digital tools and carry out actions to achieve it. An ‘agent’ can, for example, search for information, organize it, prepare an initial draft, update a system, or carry out a series of repetitive steps in an entire process. In this way, we are moving beyond simple conversational systems towards digital collaborators.
However, the increased capabilities of agentic AI also entail increased risks, bringing AI safety and the boundaries of AI autonomy to the forefront of international debates on artificial intelligence. A key priority is to ensure that these systems remain under control, with the possibility of immediate human intervention and, where deemed necessary, their shutdown. AI safety is, to a very large extent, a matter of technical safeguards and safety standards. Nevertheless, it also remains fundamentally a question of preserving human judgement and the ability to intervene. Our responsibility grows in proportion to the degree to which we entrust the autonomy of these systems. Ensuring that AI operates safely is not a constraint on technological progress, but rather a cornerstone of the trust necessary for its beneficial use.
The European approach follows in this same direction. The European Artificial Intelligence Regulation (AI Act) and Law 5321/2026, which establishes the framework for its implementation in Greece, create a common set of rules, classifying applications according to their level of risk. The principle is simple: the greater the potential impact of a system, the more stringent the requirements for safety, transparency and human oversight must be.
As agentic AI systems become part of our everyday lives, issues relating to our safe and harmonious coexistence with them will become increasingly pressing. Key priorities include the proper design of these systems in accordance with the legal framework, investment in citizens’ skills, the responsible use of data and effective human oversight. Under these conditions, agentic AI can become a new tool for growth both for the economy and for society as a whole.
At the Ministry of Digital Governance and the Special Secretariat for Artificial Intelligence and Data Governance, we play a key role in this effort. We connect public administration with research and the market, participate in shaping European policies and work towards a regulatory framework that protects citizens while simultaneously boosting innovation.