Spanish Regulator Notified of Alleged AI-Powered Data Breach

The Spanish Data Protection Agency (AEPD) has received its first report of a data breach allegedly carried out by an autonomous AI agent.

The notification describes an attack involving an agent powered by a known large language model (LLM). According to the agency, the agent autonomously searched for vulnerabilities in generic files and successfully logged into the target systems. Once access was established, the agent reportedly probed applications for further security flaws, modified personal data, and accessed financial documents, including invoices.

While the AEPD has not yet formally investigated or verified the details of the incident, the agency noted that the notification confirms that AI-related data breaches are no longer merely theoretical.

Shift in Attack Speed and Scale

The AEPD warned that although AI does not necessarily create new categories of threats, it significantly increases the speed, scale, and adaptability of cyberattacks. This shift can drastically reduce the response-time margins available to human defenders.

The agency advised that risk management models must now explicitly account for AI-assisted and AI-driven attacks. Because automation can affect the likelihood, speed, and scope of an incident, traditional response procedures designed for manual attacks may be insufficient against agents that can simultaneously analyse assets, test access methods, and adapt their behaviour.

The AEPD also emphasised the necessity of strengthening digital identity and credential security. Autonomous agents can exploit compromised accounts, API keys, or tokens with excessive permissions to navigate multiple services at machine speed, making manual intervention increasingly difficult.

Rise of Agentic Cyber Activity

This reported incident aligns with recent observations of “agentic” activity in large-scale cyber operations. There have been documented instances where AI agents were used to coordinate intrusions, scan for vulnerabilities, or perform mass credential theft across various platforms.

The AEPD clarified that even if autonomous AI is confirmed to have been used in this breach, it does not inherently mean that the underlying model or its provider’s infrastructure was compromised, or that the model was specifically designed to facilitate malicious operations.