第6期

AI审计周报 第6期

| ~

编者按：AI审计周报每周一发布，精选AI在审计、合规与监察领域的最新动态。中英文资讯混编，帮助审计人追踪行业前沿。

本期摘要

本期重点关注AI在风险管理和合规自动化领域的新进展，美国财政部发布金融机构AI风险管理框架，万事达卡推出基于交易数据的大型表格模型用于反欺诈。企业AI应用方面，多家机构强调AI治理的重要性，PwC等咨询公司加速AI转型步伐。同时，AI代理技术快速发展，OpenAI、英伟达等厂商推出企业级AI代理解决方案，但安全风险仍需谨慎管理。

AI + 金融风控 / AI in Financial Risk Management

US Treasury publishes AI risk Guidebook for financial institutions

  • 来源: AI News
  • 摘要: The US Treasury has published documents designed for the financial services sector that suggest a structured approach to managing AI risks in operations and policy. The CRI Financial Services AI Risk Management Framework (FS AI RMF) comes with a detailed Guidebook for implementation.
  • 标签: 金融监管 风险管理

Mastercard keeps tabs on fraud with new foundation model

  • 来源: AI News
  • 摘要: Mastercard has developed a large tabular model (LTM) trained on transaction data rather than text or images to address security and authenticity issues in digital payments. The foundation model is trained on billions of card transactions to enhance fraud detection capabilities.
  • 标签: 反欺诈 金融科技

Visa prepares payment systems for AI agent-initiated transactions

  • 来源: AI News
  • 摘要: Visa is testing how AI agents can initiate payments, moving beyond the traditional model where a person decides to buy something and a bank processes the transaction. This shift suggests that software agents may soon take on payment initiation roles in certain cases.
  • 标签: 支付创新 AI代理

企业AI治理与合规 / Enterprise AI Governance & Compliance

Does your organization need an AI ethics committee?

  • 来源: TechTarget Enterprise AI
  • 摘要: Ethics oversight ensures high-risk AI use is accountable, responsible and compliant. Not all organizations need a dedicated ethics committee; some can incorporate this oversight into existing risk or legal teams depending on their AI deployment scale and risk profile.
  • 标签: AI伦理 企业治理

Good governance key to reducing high AI project failure rate

  • 来源: TechTarget Enterprise AI
  • 摘要: Better governance of cutting-edge applications and the data that feeds them are key to overcoming development obstacles. Databricks exec Craig Wiley emphasized that proper governance frameworks are essential for AI project success in enterprise environments.
  • 标签: 项目管理 数据治理

EU Data Act: Time for a Reality Check

  • 来源: Corporate Compliance Insights
  • 摘要: New EU Data Act rules could spark compliance tension for organizations: share too much personal data and run afoul of GDPR, share too little and face Data Act enforcement. Companies need to carefully balance these competing regulatory requirements.
  • 标签: 数据合规 GDPR

AI代理与自动化技术 / AI Agents & Automation

NVIDIA wants enterprise AI agents safer to deploy

  • 来源: AI News
  • 摘要: The NVIDIA Agent Toolkit is Jensen Huang’s answer to enterprises asking how to deploy AI agents without losing control of data and liability. Announced at GTC 2026, it’s an open-source software stack designed to help enterprises safely implement AI agents.
  • 标签: 企业安全 AI代理

Autonomous AI adoption is on the rise, but it’s risky

  • 来源: CIO.com
  • 摘要: Two AI releases this year are prompting users to give up control and let autonomous agentic tools complete tasks on their behalf. IT leaders should be prepared to deal with the consequences as tools like Anthropic’s Claude Cowork and OpenClaw gain traction.
  • 标签: 自主AI 风险管理

OpenAI is throwing everything into building a fully automated researcher

  • 来源: MIT Technology Review
  • 摘要: OpenAI is refocusing its research efforts on building an AI researcher—a fully automated agent-based system capable of tackling large, complex problems independently. This represents a significant shift in the company’s research priorities and resource allocation.
  • 标签: 科研自动化 AI代理

合规监管动态 / Regulatory & Compliance Updates

Trump Administration Releases AI Legislative Framework

  • 来源: AI Business
  • 摘要: The administration seeks to streamline regulations at the federal level, avoiding state-by-state governance, despite potential resistance from states with their own AI regulations. This framework aims to create unified federal oversight of AI development and deployment.
  • 标签: 政策监管 联邦框架

The State of OFAC Sanctions Enforcement in 2025-26

  • 来源: Corporate Compliance Insights
  • 摘要: OFAC issued 14 enforcement actions in 2025, but the more important signal may be who got penalized rather than the number of actions. The enforcement patterns provide insights into OFAC’s priorities and approach to sanctions compliance.
  • 标签: 制裁合规 执法趋势
#AI审计 #合规科技 #内部审计