ResearchResearch paperSafety & Alignment · Agents & Tool Use · Reinforcement Learning1 source · Oct 19, 2026

PROVISO: Provenance-Informed Action Selection for Secure LLM Agents

We present PROVISO, a controller that decides which action to commit after the agent proposes an action and before it is executed.

LLM agents act on the external world through tools and third-party skills, so a single executed action can cause irreversible harm.
Existing defenses judge a proposed action without tracking whether its content comes from the user or from external sources such as tool outputs, and they judge it in isolation rather than against an available alternative, which forces a trade-off between security and utility.
PROVISO encodes trusted observations, external observations, the proposed action, and an alternative into provenance-typed memory, relates them through role-specific queries, and commits the action with the highest calibrated value learned from outcomes.
On four agent security tasks spanning the agent lifecycle, no baseline exceeds PROVISO in both security and utility.

[1]PROVISO: Provenance-Informed Action Selection for Secure LLM Agents
arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 06:13 PM
We present PROVISO, a controller that decides which action to commit after the agent proposes an action and before it is executed.
LLM agents act on the external world through tools and third-party skills, so a single executed action can cause irreversible harm.

Extractive summary: sentences quoted from the sources.

Before this