Systematic evaluation of a GenAI’s performance for an Italian airport

Use Case

Cyber security

The Challenge

Following the implementation of a chatbot based on generative AI, a major airport identified the need to assess its security. The rapid evolution of AI agents and their increasing interconnection with data repositories and business applications introduce significant risks that are often underestimated by organizations.

The Solution

The team conducted a systematic analysis of the AI architecture, designing tailored attack scenarios based on how the models are actually used. Under controlled conditions, they replicated manipulation techniques to test the integrity of the AI’s responses and the protection of sensitive data embedded in the system.

The Benefits

This initiative helped mitigate emerging risks associated with GenAI before they could be exploited, ensuring a secure interaction between the chatbot and airport users. The client gained insight into the specific vulnerabilities of language models, thereby enhancing the resilience of its digital infrastructure.

Humanativa’s role

HUMANATIVA has positioned itself at the forefront of technology by conducting one of the first AI Red Teaming exercises in the infrastructure sector. Through innovative methodologies, the Group has demonstrated how security can enable the confident adoption of artificial intelligence in critical contexts.