DTAL : THE DETERMINISTIC TRANSACTION AUTHORIZATION LAYER. Paper 1: The Necessity, Threat Environment and Operating Principles of Autonomous Transaction Authorization in Agentic AI Systems
The rapid development of autonomous artificial intelligence agents introduces acybersecurity challenge that extends beyond the protection of conventional informationsystems. AI agents can independently plan tasks, invoke tools, communicate with external services, access data and initiate operations. When multiple agents interact, their combinedactivity may also produce collective behaviour that was not explicitly specified by theirindividual operating instructions. Traditional cybersecurity controls remain essential. However, an organisation deployingautonomous agents needs an additional architectural answer to a fundamental question: Who has the final authority to authorize an external transaction initiated by thecompany's AI ecosystem? This paper proposes the Deterministic Transaction Authorization Layer (DTAL) as theexclusive authorization authority for such interactions. DTAL is designed to integrate with the ST-T1024 deterministic AI architecture and its agentic extensions A0–A16. It also incorporates A17, a research proposal for detecting emergent collective behaviour in multi-agent systems. Under the proposed arrangement, the company's AI agents operate autonomously within defined boundaries. DTAL independently evaluates and authorizes or refuses consequential external transactions according to policies established by the company. Individual human approval is not required for routine transactions. Authorization must operate at machine speed. The architecture also proposes duplicated DTAL supervision for continuity, with the CEO and CFO holding the two human governance keys for exceptional changes to the authorization authority. This paper establishes the threat environment, rationale and operating principles. Thedetailed technical architecture, protocols, interfaces, security mechanisms and verification requirements are addressed in its companion paper : Paper 2: DTAL — Architecture, Components and Authorization Protocol Keywords: Agentic AI, deterministic AI, cybersecurity, transaction authorization, multi-agent systems, collective behaviour, ST-T1024, A17, DTAL, process supervision.
Authors
- M. TAALABI
- Team LLM-DIPLOMAT
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-21
- DOI
- https://doi.org/10.5281/zenodo.24423233
- Primary Topic
- Access Control and Trust
- Type
- article
- Field-Weighted Citation Impact
- 0.00