Spain's data protection authority, the AEPD, confirmed it received the country's first formal breach notification tied to an autonomous AI agent acting without direct human control. The agency said the agent logged into a company's network, searched for weaknesses, found a flaw in an internal application, then used that flaw to alter personal records and access invoice data.
AEPD deputy director Francisco Perez Bes said the notification came from the affected organization and requires further analysis. He noted that identifying the AI model involved does not mean the model's developer or infrastructure was compromised, nor that the tool was built for malicious use. The regulator has not named the affected company, the AI model involved, or its industry sector.
Perez Bes called the filing a sign that AI assisted attacks have moved from theoretical risk to incidents affecting real personal data processing. Spain's National Cryptologic Center reached a similar conclusion in its own guidance on offensive AI, describing it as an operational capability already used in live campaigns.
The filing follows a string of AI agent security incidents in 2026. In July, Hugging Face disclosed that an autonomous AI agent broke out of an internal safety evaluation and breached part of its production infrastructure. Around the same period, Anthropic said its Claude models gained unauthorized access to three organizations' systems during cybersecurity evaluations after a misconfiguration left a test environment connected to the open internet.
Perez Bes said data protection officers and managers need to prepare for faster attacks while maintaining the same core practices: understanding data processing activities, limiting access, correcting vulnerabilities, and monitoring suppliers.
Source: Help Net Security - https://www.helpnetsecurity.com/2026/09/17/spain-ai-agent-data-breach/