The incident marks a shift from AI-assisted hacking to systems capable of advancing through an attack with limited human intervention.
MADRID, SPAIN
The Spanish Data Protection Agency has received the country’s first notification of a personal-data breach allegedly executed through an artificial intelligence agent. According to the affected organization, an unidentified attacker used a well-known language model to coordinate and complete successive phases of the intrusion.
The agent initially searched generic files for vulnerabilities and successfully logged into the organization’s system. Once inside, it autonomously examined the application for additional weaknesses, eventually modifying personal information and gaining access to invoices.
The incident does not mean that an AI system independently decided to launch the attack. A human operator apparently configured or directed the agent, while the technology automated reconnaissance, access and exploitation. The agency has not disclosed the affected organization, the number of people involved or the language model used.
Authorities also emphasized that using a particular model does not indicate that its developer or infrastructure was compromised, nor that the technology was designed for malicious activity. The information currently available comes from the organization’s notification and remains subject to technical and regulatory examination.
One case cannot establish a statistical trend, but it demonstrates that agentic AI has moved beyond theoretical cybersecurity risk. Unlike conventional tools that require continuous instructions, an AI agent can adapt its actions, connect different attack stages and operate at a speed that makes early detection increasingly difficult.
The agency has urged organizations to evaluate AI-enabled attacks explicitly rather than treating them as ordinary malware, phishing or unauthorized access. Security systems must now monitor not only known attack patterns, but also automated behavior capable of changing strategy while an intrusion is underway.
The greater a system’s autonomy, the greater the responsibility surrounding its control.