AICPA 与 AIUC 合作推出面向 AI 系统和智能体的鉴证认证
AICPA partners on new offering that assures AI systems and agents
AICPA 与 AIUC 合作推出一项新的鉴证认证,面向 AI 系统和智能体提供独立、全面的商业风险审查。该认证将 AICPA 的 SOC 2 报告与 AIUC-1 认证合并为一次由 CPA 主导的鉴证业务,AIUC-1 包含对智能体行为的实景测试,覆盖幻觉、数据泄露、操纵性请求和对抗性提示注入等风险组合。
报道 AICPA 与 AIUC 联合推出面向 AI 系统与智能体的鉴证认证,可作为国内同类业务探索方向的参照。
The AICPA and the Artificial Intelligence Underwriting Company (AIUC) have teamed up to develop an assurance certification that provides the first comprehensive, independent review of business risk related to AI systems and agents.
The joint offering delivers one CPA-led engagement that provides assurance on AI by combining a system and organization controls (SOC 2) report pioneered by the AICPA with an AIUC-1 certification. Last year, more than 40,000 organizations obtained a SOC 2 attestation or AIUC-1 certification, a news release said.
SOC 2 reports are independent examinations performed by a licensed CPA under AICPA attestation standards. The reports, according to the release, are globally recognized for demonstrating that an organization has effective controls to protect data, manage risk, and build stakeholder trust. AIUC-1 certification provides the world’s first standard for AI agent safety, security, and reliability, the release said.
“Companies have made substantial investments in AI, and investors, clients, and other stakeholders deserve a fuller, objective picture of whether the technology is safe, reliable, and properly governed,” AICPA president and CEO Mark Koziel said in the release. “This joint offering delivers that, and with CPAs leading the engagement, organizations are able to leverage the expertise, integrity, and professional judgment that builds trust across our capital markets.”
While AI is a transformative technology, the news release noted that AI misfires in critical business operations have raised safety and security concerns. Enterprises deploying AI agents need assurance that both the agents and the systems they touch are secure. AI developers must prove their agents behave as promised and meet enterprise security requirements in order to elevate trust and adoption.
“AI agents do real work, faster than anyone can check it,” Rajiv Dattani, cofounder of AIUC, said in the release. “This joint certification provides an independent answer to the two questions every organization now asks: What is our security posture? And do AI agents behave as intended? What sets AIUC-1 apart is live testing of AI agent behavior, not just a review of policies and guardrails. Agents are tested across hundreds of risk and attack combinations, from hallucinations and data leakage to manipulative requests and adversarial prompt injection.”
According to the news release, multiple CPA firms are already delivering the assurance certification to clients, and once initial engagements have concluded, a full launch is expected in early 2027.
In order to meet evolving standards for trust, security, and reliability, both SOC 2 reports and AIUC-1 certifications require annual renewal, and AIUC-1 requires quarterly technical retesting.
— To comment on this article or to suggest an idea for another article, contact Bryan Strickland at Bryan.Strickland@aicpa-cima.com.
来源:Journal of Accountancy(AICPA 会刊) · journalofaccountancy.com