We’re excited to announce that our Cresta AI Agent has earned AIUC-1 certification!
AIUC-1 is the security, safety, and reliability standard for AI agents, developed by the Artificial Intelligence Underwriting Company (AIUC). The certification process combines an independent audit of how an agent is built and governed with thousands of adversarial tests run against the live agent.
With no failures across AIUC's critical categories, Cresta AI Agent easily met the bar for certification.
Here’s why this announcement is important, and how it’s just one more step in our path to providing the most secure and reliable AI agent.
Why this matters
AI agents are moving from answering customers’ questions to taking actions on their behalf, whether that’s verifying identities, processing payments, rescheduling appointments, or resolving billing issues. That shift raises the stakes for security, compliance, and trust across the customer experience.
Traditional compliance frameworks and regulations like GDPR and SOC 2 Type 2 aren’t enough to validate an AI agent’s security in CX. This is especially true when someone is actively trying to break the agent or circumvent guardrails.
AIUC-1 closes that gap by offering:
- External evaluators: Third-party auditors reviewed our guardrails, policies, code, and logs, and AIUC ran the adversarial evaluations itself
- Real attacks: The evaluations probe for prompt injection and jailbreaks, data and PII leakage, hallucination, out-of-scope behavior, and unauthorized actions, including multi-turn conversations that adapt to the agent's replies.
- Comprehensive testing: Areas including authentication, billing, and account management are covered
Our plans to secure AI Agent even further
Our approach to security and safety will always stay ahead of AI agents’ growing threats.
Here’s what we plan on doing:
- Ongoing testing: We’ll keep running technical evaluations with AIUC every quarter and recertify annually
- Broader scope of certification: We’ll work to extend coverage to additional channels as they ship
- AI Red Team (ART): We’ve built and will continue to refine ART, a purpose-built tool that runs adversarial tests against Cresta AI Agent and scores its resilience across adversarial attack and vulnerability categories
- Continued investment in compliance. AIUC-1 sits alongside our SOC 2 Type II, ISO/IEC 42001, HIPAA, and PCI DSS-aligned controls, and we’re continuing to invest in programs such as FedRAMP
If you’d like to learn more about how Cresta keeps AI Agent secure and reliable, you can visit our Trust Center or request a demo.
