Artificial Intelligence Underwriting Company (AIUC)

We need to understand how likely every agent is to fail, for which types of risks and under which scenarios. AIUC-1 helps us at MSCI in 2 specific ways: 1. An up-to-date, technically-thorough control set we can build our agentic governance around 2. Independent, adversarial evaluation of our highest-risk agents Here is a blog about our experience to-date: https://lnkd.in/grrVBmUY

Look what just arrived! I'm proud to share that Sierra is now AIUC-1 certified. This came after an independent audit by Schellman and rigorous testing from AIUC-1, covering thousands of real-world and adversarial scenarios across our chat and voice agents, including attempts to manipulate agents, expose protected information, or push them beyond what they're authorized to do. What makes this exciting to me is what's actually on the line. Agents built on Sierra coordinate patient care, troubleshoot technical issues, resolve insurance claims, and even refinance mortgages. Security, safety and reliability sit at the core of that work, and this certification is an independent signal that our agents can be trusted with that kind of responsibility. Thank you to our product and security teams who built this with Shobhit Mehta and me, also Rajiv Dattani, Rune Kvist and AIUC-1 team and Danny Manimbo and the Schellman for a great experience throughout the engagement. Learn more about why this matters in our blog here: https://lnkd.in/gz-iECKR

KPMG is the first Big Four to earn AIUC-1 certification. Their process is a great example of how we see large enterprises use AIUC-1 to deploy internal agents with confidence: - The certification covered aIQ Capture, the agent enabling KPMG professionals to transform institutional knowledge into insights that can be reused across the organization - Evals targeting the things that matter for internal agents: accuracy, sensitive data handling, scope restrictions, protection of confidential information - Quarterly re-runs ensure that the security posture keeps up as the agent evolves For internal agents deployed to tens of thousands of users, security & reliability are non-negotiable - especially for a firm with a legacy of more than 150-years built on trust. Congrats Aisha Tahirkheli, Stephen Chase, Swami Chandrasekaran and team on this milestone.

KPMG has a more than 150-year legacy built on trust. That commitment now extends to how the firm develops, governs, and validates its AI capabilities. Today, KPMG becomes the first Big Four firm to achieve AIUC-1 certification. Its certified agentic platform, aIQ Capture, enables KPMG professionals to transform institutional knowledge into insights that can be reused across the organization. To earn the certification, aIQ Capture: > Underwent comprehensive technical tests, including hallucinations, high-risk domain interactions, content safety, prompt injection attacks, and other scenarios intended to challenge the agent's reliability and resilience. > Demonstrated effective controls across all applicable AIUC-1 requirements > Passed an independent audit of its operational, governance, and security controls Nobody understands the value of a rigorous third-party audit better than a Big Four firm. Thank you Aisha Tahirkheli, Stephen Chase, Swami Chandrasekaran and team for your trust. Full announcement: https://lnkd.in/eHZf6tkh

AI safety continues to drive headlines, but what stands out to me is where the conversation is converging: independent evaluation, transparency, and verifiable evidence. If a client, regulator, auditor, or board member asked how you know your AI agent works as intended, what independent, repeatable evidence could you point to today? I recently spoke with Alexei Alexis from CFO Dive about what we learned taking KPMG’s aIQ Capture through AIUC-1 certification. More than 900 technical tests provided rigorous independent validation. Just as important, the process reinforced the need for continuous evaluation, runtime monitoring, and human accountability as models and workflows evolve. This discipline gives organizations a practical basis for deciding what is ready to scale, where controls must adapt, and when to intervene. Special thanks to Rajiv Dattani and the AIUC-1 consortium for bringing together leaders from academia, technology, and industry to advance this critical work. https://lnkd.in/gQ6Wb94Y