Skip to content
DraconiticResearch
Search Sources API
203,948 studies31 sourcesUpdated 41 min ago
PreprintCC BY 4.0Zenodo

No Action Without Proof: A Control Plane with Verifiable Constraints, Trust Aware Coordination, and Tamper Evident Accountability

  • Zohaib Qamar1
  1. 1Universiti Kuala Lumpur, Malaysian Institute of Information Technology
Open at sourcePDFDOICite

Abstract

An agent with tools can refund the wrong customer at 3 a.m. and leave no proof of who approved it. Agents built on large language models now act. They call tools, move data, and change state. An agent that acts can fail in ways a chatbot cannot: it can hallucinate a tool argument, follow an injected instruction, or pass a false fact from one agent to another.

This paper describes a control plane that sits between a language model planner and tool execution. Every proposed action is checked against declarative policies before it runs. High stakes actions require human approval. Every decision is recorded in a tamper evident audit log that can be verified after the fact.

The work is evaluated with three open components built by the author: agent-governance, tool-truth, and gate-bypass-bench. Results reported here use only numbers verified from the public repositories on 25 October 2026.

Related

  1. The Accountability Axis for Agent Actions: A Verifiable Form (v0.1.0)
    Lunina K.
    ZenodoPreprintCC BY 4.0Full textZenodo
  2. No peace without accountability: mortality and public health in Sudan’s war
    Debarati Guha-Sapir, Thomas Jideofor Ogbu
    BMJ Global HealthPeer-reviewedCC BY-NC 4.0Full textInfectious diseaseDOAJPMC
  3. Governance, coordination, and accountability in public health emergency preparedness and response: A scoping review
    Sileshi Demelash Sasie, Melkamu Asrat Alava, Lensa Fekadu +4
    PLOS Global Public HealthPeer-reviewedCC BY 4.0Full textEpidemiology & public healthPLOS
  4. Governance-aware autonomous retail coordination in artificial intelligence cities using multi-agent reinforcement learning, blockchain accountability, and federated learning.
    Toqeer Ali Syed, Ali Akarma, Shahid Kamal +3
    Frontiers in artificial intelligencePeer-reviewedCC BYFull textEurope PMC
  5. ChunavIQ: An Open, End-to-End Verifiable Remote-Voting Prototype for India's Legal and Identity Constraints, with Deniable Revoting, Hardware-Held Trustee Keys and Adversarial Dress Rehearsals
    Ashish Ramtekkar
    ZenodoPreprintCC BY 4.0Full textZenodo

Cite

BibTeXRIS
BibTeX
@misc{qamar2026action,
  title = {{No Action Without Proof: A Control Plane with Verifiable Constraints, Trust Aware Coordination, and Tamper Evident Accountability}},
  author = {Zohaib Qamar},
  howpublished = {Zenodo},
  year = {2026},
  month = oct,
  doi = {10.5281/zenodo.23643987},
  url = {https://zenodo.org/records/23643987},
  note = {CC BY 4.0},
}
RIS
TY  - UNPB
TI  - No Action Without Proof: A Control Plane with Verifiable Constraints, Trust Aware Coordination, and Tamper Evident Accountability
AU  - Zohaib Qamar
PY  - 2026
DA  - 2026/10/25
PB  - Zenodo
DO  - 10.5281/zenodo.23643987
UR  - https://zenodo.org/records/23643987
AB  - An agent with tools can refund the wrong customer at 3 a.m. and leave no proof of who approved it. Agents built on large language models now act. They call tools, move data, and change state. An agent that acts can fail in ways a chatbot cannot: it can hallucinate a tool argument, follow an injected instruction, or pass a false fact from one agent to another.  This paper describes a control plane that sits between a language model planner and tool execution. Every proposed action is checked against declarative policies before it runs. High stakes actions require human approval. Every decision is recorded in a tamper evident audit log that can be verified after the fact.  The work is evaluated with three open components built by the author: agent-governance, tool-truth, and gate-bypass-bench. Results reported here use only numbers verified from the public repositories on 25 October 2026.
ER  -