No external source link is registered for this Q&A.
What you may be facing
A multinational trading enterprise introduces an AI agent with reasoning and planning capabilities to autonomously orchestrate and optimize end-to-end cross-border B2B payments and treasury operations. During the automatic execution of a large-value payment, significant foreign exchange losses occurred due to the AI's autonomous decision-making regarding inappropriate timing for currency conversion. Subsequently, regulatory authorities questioned the transaction for potential anti-money laundering compliance issues. The enterprise needs to clarify the ultimate parties bearing commercial risks from foreign exchange losses and administrative/civil compliance liabilities in such novel technology application scenarios.
Why this needs attention
According to industry white papers released during the 2026 World Artificial Intelligence Conference and relevant legal practices, AI agents themselves do not possess civil subject status and cannot independently bear legal liability. The current legal framework follows the "whoever uses it bears responsibility" principle. As the deployer, authorizer, and beneficiary of the AI, the enterprise bears primary responsibility for the authenticity of the transaction background, foreign exchange reporting, and data compliance. Clarifying liability attribution helps enterprises effectively isolate and manage potential legal and financial risks through contractual agreements and technical means while enjoying the efficiency of automation.
What you can do now
- Review and improve the Service Level Agreement (SLA) with the AI service provider, clearly stipulating compensation mechanisms for losses caused by algorithmic defects, model vulnerabilities, or failure to execute instructions.
- Clearly specify the allocation of foreign exchange loss burdens and telegraphic transfer fee modes (such as OUR/SHA/BEN) in cross-border trade contracts, establishing rules for distributing exchange rate fluctuation risks.
- Establish complete records of the AI authorization chain, retaining machine-readable parameters for scope of authority, transaction limits, and compliance rules to ensure traceability for regulatory reviews.
- Set manual review thresholds, mandating human confirmation for large-value payments or transactions involving high-risk countries to avoid complete reliance on autonomous AI decision-making.
- Deploy an independent compliance verification layer to conduct secondary screening of payment instructions initiated by the AI, ensuring compliance with Anti-Money Laundering (AML), sanctions screening, and foreign exchange management regulations.
This Q&A is compiled based on the report by Economic Observer on July 19, 2026, regarding AI agents reshaping cross-border payments, as well as general legal principles. AI agents do not possess independent legal personality; legal liabilities arising from their actions are attributed to the human entities behind them (enterprises, service providers, etc.). Specific liability allocation requires comprehensive judgment based on service contract terms, the degree of fault of each party, and specific regulatory requirements of the relevant jurisdiction (such as China's foreign exchange management regulations and the EU's MiCA framework). This article does not constitute formal legal advice; please consult professional lawyers for specific cases.
