In recent days, concern about the advancement of artificial intelligence (AI) has intensified. The debate now focuses not only on generative tools like ChatGPT, which produce content from human instructions, but on agentic AI, which executes tasks and makes decisions on its own.
What is agentic AI?
Researcher Alicia Ortiz de Zárate from the Autonomous University of Madrid explains that these systems have greater capacity for action, complexity, and autonomy. “Unlike a tool that simply responds to a request, an artificial intelligence agent can execute multiple actions to complete a task,” she notes.
Risks and responsibilities in critical infrastructure
Nuria Oliver, director of the ELLIS Alicante Foundation, warns about the safety of essential sectors such as hospitals, energy systems, and defense. According to her, the known risks of traditional AI could increase with the autonomy of these agents. “The more strategic the areas in which they operate, the greater the risk and the harder it may be to determine responsibilities,” she said.
Oliver questions the idea that AI “got out of control.” In her view, responsibility should fall on designers and developers. She compared the situation to other sectors where controls are applied before a product reaches the market and advocates for a similar logic for AI.
International governance and control
Ramón López de Mántaras, founder of the Institute of Artificial Intelligence Research at CSIC, argues that talking about a “rebellion” of agents can shift responsibility from developers to the technology itself. Rather than halting development, he proposes steering AI toward specific, transparent, robust, and controllable tools.
Senén Barro, director of the Singular Center for Research in Intelligent Technologies at the University of Santiago de Compostela, highlights the need for evaluation mechanisms and clarity about who is accountable for agents’ actions.
The European Union already has an AI regulation, but experts warn that its operation will need to be examined in light of the arrival of more autonomous agents. López de Mántaras calls for international governance that establishes clear responsibilities, which would require agreements among technological powers such as China and the United States.
Thus, the debate no longer revolves solely around AI’s capabilities, but also about who controls its decisions, who is accountable when something goes wrong, and what limits should be set before these systems access increasingly sensitive areas.
Source credits: deultimominuto.com
