Yubico partners with IBM and Auth0 to secure agentic AI and harness its power while ensuring trust and governance

Yubico has announced a strategic collaboration with IBM and Auth0 to help organisations securely deploy AI agents at scale. It introduces a new model of human-in-the-loop authorisation for high-risk automated actions.

The Rise of Agentic AI and Its Risks

As enterprises rapidly adopt AI-driven automation, agentic AI systems are moving beyond advisory roles. They now act as “digital workers” that execute transactions, deploy code and access sensitive systems.

While this shift promises productivity gains, it also introduces governance and security challenges. As a result, organisations must rethink control models.

According to recent industry data, 87 per cent of organisations report increased risk from AI vulnerabilities. Meanwhile, 77 per cent lack foundational AI security practices.

Why Traditional Access Controls Are Failing AI

Traditional identity and access models were built for human actions. However, AI agents now operate autonomously. Because of this, organisations must prove not only that an action occurred, but also who approved it.

The joint solution from Yubico, IBM, and Auth0 addresses this risk. Specifically, it requires cryptographically verified human approval for critical AI actions.

Introducing Human-in-the-Loop Authorisation

Traditional identity and access management models were designed for human-initiated actions. However, as AI agents begin operating autonomously, organisations must be able to prove not only that an action occurred, but also who authorised it.

The new architecture introduces a Human-in-the-Loop model that enables AI systems to operate autonomously for routine tasks, while escalating high-risk actions for secure human approval.

This approach is designed to support a wide range of enterprise use cases, including:

• High-value financial transactions
• Deployment of production code
• Access to sensitive corporate data
• Procurement approvals and vendor onboarding
• Security response actions impacting system availability.

How Hardware-Backed Authentication Secures AI Decisions

At the centre of this model is Yubico’s hardware-backed authentication. The YubiKey ties approvals to a verified individual and physical presence.

When a high-risk action occurs, Auth0 initiates a secure approval request. At the same time, IBM’s AI agents orchestrate the workflow. The authorised user must then authenticate with a YubiKey.

This process provides cryptographic proof that:

• The correct individual approved the action
• The approval cannot be replayed or intercepted
• The decision is securely bound to a verified identity.

Building Trust and Auditability in AI Workflows

As a result, organisations gain a clear audit trail. They also achieve non-repudiation for critical decisions. This supports compliance, financial accountability, and risk management requirements.

Scaling Secure AI Across the Enterprise

According to Sheryl Chamberlain, Vice President at Yubico, organisations are reaching a critical inflection point. AI is moving from experimentation to execution.

AI agents are becoming part of the enterprise workforce. They can now take real action across systems. However, organisations must ensure that speed does not undermine trust or governance.

By combining hardware-backed authentication with identity orchestration, organisations can scale AI securely. At the same time, they keep humans in control of critical decisions.

Inside the Yubico, IBM, and Auth0 Architecture

The collaboration brings together:

• IBM’s WatsonX AI capabilities to analyse data and execute workflows
• Auth0’s identity orchestration to manage secure approval flows
• Yubico’s hardware root of trust to enforce human verification.

A New Governance Model for Autonomous Systems

Together, the solution enables organisations to bridge the gap between autonomy and accountability, ensuring that AI can operate at machine speed without compromising security.

As AI adoption accelerates across industries, including financial services, software development, and procurement, organisations must rethink how they govern automated decision-making.

The Human-in-the-Loop model represents a shift from reactive security controls to proactive, policy-driven authorisation embedded directly into AI workflows.

By integrating cryptographically enforced human approval into high-risk actions, organisations can unlock the full potential of AI-driven automation while maintaining control, transparency, and trust.

About Cyber News Live

Stay ahead with Cyber News Live! First, we deliver real-time reporting and sharp threat intelligence. Additionally, we provide educational content for professionals, practitioners, and curious minds. From there, whether it’s breaking breach alerts or deep dives into attack vectors, we cover it all. Ultimately, our mission is clear: we make complex cyber topics understandable. And beyond that, we ensure critical knowledge stays accessible to everyone.