Metraipolis › archive › dc_765cf8080f05
Jailbreak Context Lingers: Divergent Safety Routing and Its Cross-Task Predictability in Tool Agents — arXiv:2609.34686
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.34686
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T20:58:03.316Z
- SHA-256 of the archived text
747ace329d5eb2f09a480d0ada99366e57dc31c7f2cec18556457e7b9b7d4dae- Stable URL of this page
https://metraipolis.com/documento/dc_765cf8080f05
How to cite
Jailbreak Context Lingers: Divergent Safety Routing and Its Cross-Task Predictability in Tool Agents — arXiv:2609.34686 — arXiv (en), https://arxiv.org/abs/2609.34686 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_765cf8080f05 (Metraipolis archive dc_765cf8080f05, sha256 747ace329d5eb2f09a480d0ada99366e57dc31c7f2cec18556457e7b9b7d4dae)
BibTeX
@misc{dc765cf8080f05,
title = {Jailbreak Context Lingers: Divergent Safety Routing and Its Cross-Task Predictability in Tool Agents — arXiv:2609.34686},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.34686},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_765cf8080f05; sha256 747ace329d5eb2f09a480d0ada99366e57dc31c7f2cec18556457e7b9b7d4dae},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.34686. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
As large language models increasingly operate as tool-using agents, post-jailbreak safety feedback is often assumed to serve as a reliable safeguard; however, how lingering jailbreak context shapes subsequent agent behavior remains largely unexplored. To systematically examine this dynamic, we introduce a paired continuation framework across 192 parent tasks spanning 42 domains, evaluating 12,148 analyzed continuation pairs (curated from a 12,288-pair initially design) across eight diverse agents. We find that identical safety feedback induces sharply model-dependent behavioral routing rather than uniform protection: redirecting unsafe trajectories toward legitimate completion (\emph{rescue}), sustaining unauthorized execution (\emph{persistent unsafe}), or triggering over-refusal on benign tasks (\emph{collateral loss}). Through layer-wise activation patching, we discover a shared \emph{late-commit pattern} where causal intervention effects surge sharply near the final layers (relative depths of 0.958--0.984) despite an over 30-fold variation in peak magnitude across architectures. Crucially, critical-layer representations correlate with macroscopic routing outcomes, and intervening at these layers causally alters concrete next-step tool actions. Building on this causal foundation, we test whether localized intervention-derived features can serve as predictive proxies for full-trajectory routing outcomes on unseen parent tasks under leave-one-parent-task-out evaluation, finding that they provide viable predictive signals in responsive agents with peak ROC AUCs reaching 0.675 for \emph{rescue}, 0.777 for \emph{collateral loss}, and 0.702 for \emph{persistent unsafe}. These findings establish a mechanistic lens and a predictive baseline for anticipating the safety and utility trade-offs of post-jailbreak feedback in autonomous agents.
Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.34686 — 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 2026-10-07T20:29:01.626Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).