Spain's Data Protection Agency (AEPD) published a notice on September 14 describing what it calls the first personal data breach notification in which the attacker was an artificial intelligence agent operating autonomously: it found vulnerabilities, logged in, and modified personal data and invoices without receiving step-by-step instructions from a human.
According to the AEPD notice, the attacking agent "started a search for vulnerabilities in generic files, and successfully logged in" to the affected organization's application. Once inside, it "began autonomously searching for vulnerabilities in the application," which ultimately allowed it to modify personal data and access invoices. The agency does not identify the affected organization or the AI model used by the attacker — standard practice for this type of notification.
What sets this incident apart from a conventional automated attack (a script or a bot running fixed rules) is that the agent planned intermediate tasks, used tools, and adjusted its behavior based on what it discovered along the way, without needing detailed instructions for each step. It's the same "plan, act, observe, adjust" loop that makes AI agents useful for legitimate work — applied here to intrusion and vulnerability exploitation.
The notice lays out concrete recommendations: explicitly factor AI-assisted or AI-executed attacks into risk analyses for every data processing activity, review threat response times on the assumption that attackers can now move faster than a human, strengthen digital identity and credential controls, and deploy automated detection and response instead of relying solely on manual oversight — without abandoning the classic fundamentals of data minimization, access control, and vendor oversight.
Carlos Montiel is an enterprise AI solutions architect. He implements LLMs, Agents, RAG, and orchestrators for businesses in Guatemala and Latin America. Reach out for a consultation.
Contact Carlos Montiel