A Proposal for Evaluating the Operational Risk for ChatBots based on Large Language Models

The emergence of Generative AI (Gen AI) and Large Language Models (LLMs) has enabled more advanced chatbots capable of human-like interactions. However, these conversational agents introduce a broader set of operational risks that extend beyond traditional cybersecurity considerations. In this work, we propose a novel, instrumented risk-assessment metric that simultaneously evaluates potential threats to three key stakeholders: the service-providing organization, end users, and third parties. Our approach incorporates the technical complexity required to induce erroneous behaviors in the chatbot--ranging from non-induced failures to advanced prompt-injection attacks--as well as contextual factors such as the target industry, user age range, and vulnerability severity. To validate our metric, we leverage Garak, an open-source framework for LLM vulnerability testing. We further enhance Garak to capture a variety of threat vectors (e.g., misinformation, code hallucinations, social engineering, and malicious code generation). Our methodology is demonstrated in a scenario involving chatbots that employ retrieval-augmented generation (RAG), showing how the aggregated risk scores guide both short-term mitigation and longer-term improvements in model design and deployment. The results underscore the importance of multi-dimensional risk assessments in operationalizing secure, reliable AI-driven conversational systems.

A secondary immune response based on co-evolutive populations of agents for anomaly detection and characterization

The detection of anomalies in unknown environments is a problem that has been approached from different perspectives with variable results. Artificial Immune Systems (AIS) present particularly advantageous characteristics for the detection of such anomalies. This research is based on an existing detector model, named Artificial Bioindicators System (ABS) which identifies and solves its main weaknesses. An ABS based anomaly classifier model is presented, incorporating elements of the AIS. In this way, a new model (R-ABS) is developed which includes the advantageous capabilities of an ABS plus the reactive capabilities of an AIS to overcome its weaknesses and disadvantages. The RABS model was tested using the well-known DARPA'98 dataset, plus a dataset built to carry out a greater number of experiments. The performance of the RABS model was compared to the performance of the ABS model based on classical sensitivity and specificity metrics, plus a response time metric to illustrate the rapid response of R-ABS relative to ABS. The results showed a better performance of R-ABS, especially in terms of detection time.