📄 中文摘要
在过去六个月中，企业人工智能团队中出现了相似的模式。A2A和ACP在架构评审中引起了极大的关注，协议设计优雅，演示效果令人印象深刻。然而，三周后，当有人询问“凌晨2点，哪个代理授权了那笔5万美元的供应商付款？”时，兴奋的气氛转变为担忧。这一现象揭示了在快速发展的AI技术应用中，治理和授权机制的缺失可能导致的风险和挑战。
人工智能、A2A与治理缺口
出处: AI, A2A, and the Governance Gap
发布: 2026年2月17日
📄 English Summary
AI, A2A, and the Governance Gap
Over the past six months, a recurring pattern has emerged within enterprise AI teams. A2A and ACP protocols have garnered significant attention during architecture reviews due to their elegant design and impressive demonstrations. However, three weeks into production, concerns arise when questions about authorization for significant transactions, such as a $50,000 vendor payment at 2 AM, surface. This shift from excitement to concern highlights the potential risks and challenges stemming from a governance gap in the rapidly evolving landscape of AI technology applications.