#KEYMASTER: Agentic AI Risks and The Role of Identity

In this episode of #KEYMASTERs, Sven Rajala, International PKI Man of Mystery, is joined by Ryan Sanders, Senior Director of Product and Customer Marketing at Keyfactor, to unpack one of today’s most discussed topics: Agentic AI.

They begin by defining the concept; while generative AI creates content in response to prompts, agentic AI takes things a step further by acting independently, making decisions, and executing tasks across systems. This shift from generating to acting introduces powerful capabilities, but also serious security implications.

Ryan points out that agentic AI agents often need direct access to enterprise systems and sensitive applications to perform useful work. That means trust and authentication become central concerns: who or what is accessing a system, and how do we verify it securely?

Watch video on YouTube

Most agentic AI projects today are still in design or pilot stages, largely because organizations do not yet fully trust these systems. Many rely on static credentials like user or service accounts, which create significant risk. Instead, Ryan and Sven discuss moving toward stronger, scalable authentication methods such as PKI (Public Key Infrastructure) and OAuth frameworks, using certificates as credentials to establish identity, issue short-lived tokens, and enable secure communication through mutual TLS (mTLS).

The discussion also explores how these agents are likely to run in containerized and service mesh environments, making it critical to apply existing best practices from DevSecOps, such as container signing, provenance attestation, and policy enforcement, to this new class of workloads.

As the technology evolves toward autonomous and multi-agent systems, the challenge will be to combine this agility with enterprise-grade trust. The session concludes by highlighting the growing ecosystem of emerging standards from OWASP, the Cloud Security Alliance, and Keyfactor’s own recent white paper on securing AI agents.

Key takeaways

• Agentic AI represents the next evolution of AI,  from generating content to taking autonomous action.
• The biggest barriers to adoption today are trust and secure authentication.
• Static credentials (user/service accounts) are risky and easily exploited.
• PKI and OAuth offer a more secure and scalable approach to authenticating AI agents.
• Certificates are the strongest form of credentials, verifiable, rotatable, and tied to a root of trust.
• Service meshes, container signing, and attestation will play a key role in securing agentic AI deployments.
• Standards bodies like OWASP and the Cloud Security Alliance are developing frameworks to guide safe adoption.

Learn more

Read hour whitepaper: Securing Agentic AI with Zero Trust

Find all our #KEYMASTER videos and more on Keyfactor for Developers – Your hub for cryptography, PKI, and signing.