Verifiable Agentic Commerce: A Protocol-Driven Framework for Strategic Information Disclosure and Hallucination-Resistant Autonomous Negotiation

Engineering notes

Engineering notes will be added by the aipentium editorial team.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为大语言模型、生成式AI、ChatGPT相关技术、计算机视觉、深度学习等高价值论文补充中文说明。

Original abstract

As Artificial Intelligence transitions from a passive retrieval tool to an autonomous economic actor, the emergence of Agentic Commerce necessitates a robust infrastructure for secure bilateral negotiation. While current Large Language Model (LLM) agents demonstrate advanced reasoning capabilities, they present two critical vulnerabilities that impede commercial adoption: strategic information leakage, defined as the unintentional disclosure of user budgets or constraints, and generative hallucination, which refers to the fabrication of non-binding or invalid contractual terms. This work proposes a novel Bilateral Information-Gated Protocol (BIGP) to standardize interactions between Buyer and Seller agents. In contrast to public-scraping models, this protocol employs a Private-API Gateway, enabling the Seller Agent to serve as a controlled intermediary to the merchant’s internal inventory and pricing data. To promote fair practices, the framework incorporates a Deception Detection Layer that utilizes cryptographic oracles to verify scarcity claims without disclosing proprietary information. To mitigate the risks associated with LLM hallucinations, this work presents the Deterministic Settlement Handshake. This mechanism separates the generative negotiation phase, or semantic haggling, from the transactional execution phase. All outcomes negotiated by LLMs must pass through a non-generative, symbolic validation layer, which is verified against a cryptographically signed Human-in-the-Loop Constraint Manifest. This process ensures that no agent can commit a user to an agreement that breaches predefined budget or inventory constraints. Experimental evaluation will assess the protocol’s ability to maintain a Pareto-optimal frontier while minimizing information entropy leakage. The expected outcomes seek to establish a foundational blueprint for a trustless, standardized, and legally binding agentic marketplace.

Links and sources

Comments