Introducing the world’s first AI agent standard: AIUC-1 - 'SOC 2 for AI agents'. AIUC-1 accelerates adoption of secure AI agents by helping businesses answer the question: “can we trust this AI agent?” Executives face a difficult choice: get left behind by competitors adopting AI, or move fast and risk making headlines if an AI agent malfunctions. We asked more than 500 industry leaders about their concerns with AI. Their answers were remarkably consistent: — Data leakages — Jailbreaks — IP concerns — Hallucinations — Brand disasters The AIUC-1 standard for AI agents covers all core enterprise concerns in six principles: 1. Data & privacy 2. Security 3. Safety 4. Reliability 5. Accountability 6. Society The principles cover 51 testable requirements backed by 187 controls — at the technical, operational, and legal levels — so leaders can move fast with confidence. AIUC-1 was developed alongside experts in AI research, cybersecurity, legal, audit, compliance, including Christina Liaghati, PhD at MITRE, Phil Venables at Google Cloud, Lena Smart, Hyrum Anderson at Cisco, Sanmi Koyejo at Stanford University, John Bautista at Orrick, Herrington & Sutcliffe LLP, Jim Reavis at Cloud Security Alliance. Our founding team (Rajiv Dattani Brandon Kent Wang) combines AI expertise from Anthropic, the Center for AI Safety, and METR’s frontier model evaluations, with institutional risk experience from McKinsey's insurance practice, and an underwriting startup. We are already helping leading AI companies, including Cognition (code generation), ada (customer support), Alpharun (AI recruiter) prove their security and speed up secure deployments.
Important step forward in ensuring AI agents are safely adopted by enterprises. Excited to see this go live.
Looks really interesting, CSA will take a look at possibly mapping this to our AI Controls Matrix.
Proud of Ada for all the security, safety and reliability we built into our AI Agent platform that set us up to be certified by AIUC-1. Congrats on the launch Rune Kvist, Brandon Kent Wang!
Loving this working principle from the AIUC founders. Really looking forward to seeing how an active understanding from field experts can lead the insurance market in quantifying product validity and henceforth risk transfer in a rapidly evolving AI space. Well done to Rune, Roger and Brandon. Game changing move in unlocking consumer comfort and the inevitable adoption of the most exciting of technologies?
It has been a pleasure to develop AIUC-1 over the past few months. So grateful to our partners and collaborators across AI, security, enterprise, and insurance to make sure AIUC-1 addresses the AI risks that matter
Congratulations to Rune Kvist and the whole team behind AIUC-1. The world needs more big thinking about how to achieve a positive future for humanity + AI
Our MITRE ATLAS team is excited to continue collaborating with you all!
So excited to introduce ourselves to the world! Builders I've spoken to know this is a big deal
This is freaking amazing! Well done Rune and team! 😍
SOC 2 Auditor | Cybersecurity Compliance
3moAre there organizations that are certified/accredited to perform assessments and issue certificates? I didn’t see anything on the website that talks about the process for becoming certified and who the gatekeepers are for the certification.