Symbolic Guardrails for Domain-Specific Agents: Stronger Safety and Security Guarantees Without Sacrificing Utility

Dimensions overall score 8.0

• ing agent benchmarks, which cannot be guaranteed by symbolic guardrails, and what alternative approaches are available?claim 82%
• security, and utility of AI agents? To answer these questions, we conduct a three-part study. First, we perform a systematic literature review to collect 80 state-of-the- art AI agent safety and security benchmarks and analyze the safety and security policies they evaluate. Secondclaim 82%
• marks evaluate agent safety and security. Specifically, through a systematic literature review, we analyze a large corpus of AI agent benchmarks that evaluate safety or security, and we extract and classify the safety or security policies they define. 3.1.1 Identifying Benchmark Papersclaim 82%
• and Security Guarantees Without Sacrificing Utility Yining Hong Carnegie Mellon University Pittsburgh, PA, USA yhong3@andrew.cmu.edu Yining She Carnegie Mellon University Pittsburgh, PA, USA yiningsh@andrew.cmu.edu Eunsuk Kang Carnegie Mellon University Pittsburgh, PA, USA eunsukk@andrew.cmu.edu Christopher Sclaim 82%
• isting agent benchmarks, which can be guaranteed by symbolic guardrails, and by what mechanisms?claim 82%
• we limit the search period to January 1, 2022 through March 1, 2026. Following prior recommendations [20], we first define a seed set of 15 relevant benchmark papers with which we were already familiar. By analyzing this seed setclaim 82%

Competitive landscape

No public competitor map is available for this paper yet.