Search papers, labs, and topics across Lattice.
This paper introduces OpenART, a scalable framework for agent red teaming that utilizes environment evolution to address cumulative risks in AI safety assessments. By providing over 10,000 validated stateful scenarios and employing the Evolutionary Markov Hypergraph Attack (EMHA), the authors demonstrate that environment evolution significantly enhances the detection of safety failures, achieving an 85.0% Attack Success Rate across various agent-model configurations. Notably, the advantage of EMHA over traditional instruction-only evolution grows with task complexity, highlighting the importance of dynamic environments in evaluating agent behavior.
Environment evolution can reveal 17% more safety failures in complex tasks compared to static benchmarks, reshaping our understanding of agent vulnerabilities.
AI agents operate in persistent environments where early state changes can influence decisions far into the future. Unlike conventional language-model interactions, agent behavior is mediated through a shared state that is repeatedly modified and reused across long-horizon workflows. Current safety benchmarks often fail to capture these cumulative risks because they focus on short, static tasks. To address these limitations, we introduce OpenART, an open-ended arena for scalable agent red teaming through environment evolution. OpenART provides over 10,000 validated stateful scenarios across 50 domains, drawing from a pool of more than 500,000 tools and skills. These tasks require a median of 97 tool calls and enable unified evaluation across 75 different agent-model configurations. To systematically explore these evolving attack surfaces, we propose the Evolutionary Markov Hypergraph Attack (EMHA). EMHA is a black-box policy that performs feedback-driven environment evolution by coordinating authorized state transitions without requiring parameter updates. Throughout the evaluation, task objectives remain fixed while only the environment state changes. Across all configurations, EMHA achieves a pooled Attack Success Rate (ASR) of 85.0%. Its advantage over instruction-only evolution increases from approximately 2% on simple environments to over 17% on the most complex ones, demonstrating that environment evolution increasingly exposes safety failures as task complexity grows. Furthermore, our analysis shows that the specific runtime implementation of an agent explains a significant portion of safety variation beyond the underlying model's capabilities. These results establish OpenART as a scalable foundation for studying agent safety in complex, evolving environments.
