Spain’s data protection authority has received what it describes as its first notification of a personal-data breach allegedly executed by an AI agent, offering a rare documented example of an autonomous system carrying out multiple stages of a cyberattack with limited human intervention.
The Spanish Data Protection Agency (AEPD) disclosed the case on September 14. According to the organization that reported the breach, an AI agent powered by an unnamed but widely known large language model (LLM) first logged into the targeted system and then autonomously searched for weaknesses in the application. After identifying a vulnerability, the agent allegedly modified personal information and accessed billing records.
Neither the affected organization nor the AI model has been identified. The AEPD also stressed an important distinction: using a particular model in an attack does not mean the model itself or its provider’s infrastructure was compromised, nor does it indicate that the technology was created for malicious purposes. The regulator is still reviewing the notification, so the incident should not yet be treated as a fully established account of an autonomous AI attack.
The significance lies less in a new hacking technique than in automation. The AEPD says AI does not fundamentally create new cyber threats, but can dramatically increase their speed, scale and adaptability. An agent capable of discovering weaknesses, selecting actions and executing them with little supervision can compress several steps that traditionally required sustained human involvement.
One case is not evidence of a broad trend. But for companies handling personal information, it suggests incident-response times may need to shrink as autonomous systems become more capable of chaining conventional attack techniques together.