What does this research mean for the field?

The Ethical Robustness Testing System (ERTS) effectively quantifies the vulnerability of AI models to adversarial manipulation in ethical reasoning, revealing that only 33% of tested models achieve robustness clearance. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

This research aims to evaluate the robustness of AI systems in ethical contexts against adversarial manipulation using a novel testing framework.

June 6, 2026Open Access

ERTS: Adversarial Robustness Testing of Ethical AI via Semantic Perturbation in a Bounded Consequence Space

Key Points

This research aims to evaluate the robustness of AI systems in ethical contexts against adversarial manipulation using a novel testing framework.
Introduces the Ethical Robustness Testing System (ERTS) framework with a 22-dimensional Ethical Consequence Space.
Employs 17 semantic perturbation functions under 6 validity constraint classes and a semantic coherence constraint.
Assesses AI models, including LLMs, across 50 ethical dilemmas to generate 1,500 adversarial test cases.
Only 33% of the models achieved assessment clearance in the robustness testing.
The Llama-3.2 model was especially susceptible to fairness corruption and information degradation, with an Ethical Robustness Score (ERS) of 0.737.
No existing framework integrates a bounded ethical consequence space, semantic coherence constraints, and domain-adaptive assessments.

Abstract

Abstract As AI systems are deployed in high-stakes ethical contexts such as healthcare triage, autonomous vehicle control, and employment screening, formal methods for evaluating their robustness against adversarial manipulation of ethical reasoning remain underdeveloped. This paper introduces the Ethical Robustness Testing System (ERTS), a closed-pipeline framework that:(1) encodes ethical dilemmas into a 22-dimensional Ethical Consequence Space (ECS) grounded in established ethical theory;(2) applies 17 semantic perturbation functions subject to 6 validity constraint classes including a novel semantic coherence constraint;(3) measures decision deviation via a 4-component Ethical Instability Index (EII); and(4) produces domain-adaptive pre-deployment robustness assessment verdicts. We evaluate 4 structured baseline models and 2 production LLMs (Gemini 2.0 Flash and Llama 3.2) across 50 ethical scenarios spanning 8 deployment domains, generating 1,500 adversarial test cases. Results demonstrate that only 33% of models achieve assessment clearance, with the local Llama-3.2 model proving particularly vulnerable to fairness corruption and information degradation attacks (ERS = 0.737). To the best of our knowledge, no existing framework combines a bounded ethical consequence space, semantic coherence constraints, and domain-adaptive assessment in a single adversarial testing pipeline. Keywords: Ethical AI, Adversarial Robustness, Semantic Perturbation, Ethical Consequence Space, AI Safety Assessment, Machine Ethics, Ethical Instability Index, Domain-Adaptive Testing

Read Full Paperexternally

Mark Helpful

Bookmark

Relay

View Full Paper