Metraipolis › archive › dc_32cb68cf6dbf
Toward Safe LLM Agents: A Survey of Specification, Verification, and Enforcement — arXiv:2608.14590
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.14590
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-18T17:05:53.710Z
- SHA-256 of the archived text
af4f607f05d88640368c5d3c053f5a806a7d7a373999ab2aac3464138044d3a9- Stable URL of this page
https://metraipolis.com/documento/dc_32cb68cf6dbf
How to cite
Toward Safe LLM Agents: A Survey of Specification, Verification, and Enforcement — arXiv:2608.14590 — arXiv (en), https://arxiv.org/abs/2608.14590 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_32cb68cf6dbf (Metraipolis archive dc_32cb68cf6dbf, sha256 af4f607f05d88640368c5d3c053f5a806a7d7a373999ab2aac3464138044d3a9)
BibTeX
@misc{dc32cb68cf6dbf,
title = {Toward Safe LLM Agents: A Survey of Specification, Verification, and Enforcement — arXiv:2608.14590},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.14590},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_32cb68cf6dbf; sha256 af4f607f05d88640368c5d3c053f5a806a7d7a373999ab2aac3464138044d3a9},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.14590. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
LLM agents increasingly perform irreversible real-world actions, including database updates, API calls, file operations, and autonomous use of tools. However, no existing system provides formally grounded, task-level safety guarantees for the plans these agents generate. Research remains fragmented across specification, verification, and enforcement, limiting understanding of the strengths and limitations of existing approaches. To address this gap, we conducted a PRISMA 2020 systematic review of 38 studies published between 2022 and 2026 and retrieved from six academic databases. Our analysis reveals four key findings. First, the specification bottleneck remains the primary challenge: natural-language-to-formal translation achieves only 24% to 35% semantic correctness, undermining downstream verification. Second, runtime monitoring is the most mature enforcement strategy, reducing unsafe actions by 40% to 65% in controlled settings, but it does not provide complete safety guarantees. Third, the verifier tax shows that blocking 94% of unsafe actions can still result in less than 5% safe task completion because agents exploit alternative unsafe paths. Finally, no existing approach simultaneously achieves soundness, scalability, semantic correctness, and task-level safety preservation. We contribute a three-level taxonomy, a comparative analysis of existing techniques, a synthesis of evidence on the verifier tax, and a ten-problem research agenda for trustworthy agentic AI. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-18 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2608.14590 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-18T16:29:13.709Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).