PKI · Artificial intelligence · Governance
When AI Meets Identity:
Why PKI Thinking Is Becoming Essential for AI Governance
Artificial Intelligence is introducing a challenge that security professionals have encountered before—only in a different form.
The question is no longer simply whether a user should be trusted to access a system. Increasingly, organisations must determine whether an autonomous or semi-autonomous AI system should be trusted to act on their behalf.
For professionals working in Public Key Infrastructure (PKI), this problem is surprisingly familiar.
For years, PKI architects have designed and operated systems that establish trust through digital identities, certificates, cryptographic keys, certificate authorities, issuance policies, revocation mechanisms, and lifecycle management. Although AI agents do not necessarily rely on X.509 certificates or traditional PKI for every interaction, they introduce many of the same governance challenges.
The technology may be different, but the underlying security questions remain remarkably consistent.
The Identity Challenge
Modern AI-enabled workflows rarely operate in isolation. Instead, they interact with a wide range of services and platforms, often using credentials that already exist within an organisation.
These workflows may rely on:
- API keys
- OAuth access tokens
- Service accounts
- Secrets stored within automation platforms
- Model access credentials
- Permissions to invoke external tools
- Retrieval systems containing enterprise data
- Human approval workflows
It is tempting to ask whether these technologies “use PKI.” However, that is the wrong question.
“What identity is performing this action, what authority has been granted to it, and how is that authority governed throughout its lifecycle?”
This subtle shift changes the discussion from technology to trust.
Why Governance Matters
Recent industry guidance has highlighted that AI security risks extend far beyond prompt injection or hallucinations.
Among the risks identified in the OWASP Top 10 for Large Language Model Applications are sensitive information disclosure, supply chain compromise, improper output handling, and excessive agency.
Excessive agency deserves particular attention.
As AI systems become capable of reading enterprise data, creating support tickets, writing files, sending emails, executing code, or triggering automated workflows, their credentials effectively become part of the organisation’s security boundary.
If those identities are poorly governed, AI can unintentionally inherit privileges far beyond what was originally intended.
Applying PKI Principles
This is where PKI experience becomes valuable.
PKI has always required practitioners to think in terms of identity lifecycle rather than static credentials. The same mindset applies equally well to AI governance.
Every identity should be:
- Issued with an appropriate level of authority.
- Granted only the permissions required to perform its intended function.
- Continuously monitored.
- Rotated where appropriate.
- Revoked when no longer required.
- Audited throughout its lifetime.
These principles naturally support broader security objectives, including least privilege, credential rotation, approval workflows for high-risk actions, comprehensive logging, traceability, and clear separation between user intent and automated execution.
Rather than treating AI identities as exceptional cases, organisations can often strengthen their governance by applying established identity management disciplines.
A Practical Starting Point
One effective exercise is to examine a single AI-assisted workflow. Identify every credential that the workflow can access or use. Then classify each credential according to ownership, lifetime, and privilege level:
- Human-owned
- Service-owned
- Shared
- Long-lived
- Short-lived
- Privileged
- Unknown
Credentials that are shared, long-lived, highly privileged, or lack clear ownership should be considered priorities for review.
This simple inventory frequently reveals governance gaps that would otherwise remain hidden.
Looking Ahead
AI is not replacing the principles of identity security—it is expanding where those principles must be applied.
As organisations adopt increasingly capable AI systems, identity, authority, and credential lifecycle management will become foundational elements of AI governance.
The technologies may evolve rapidly, but the underlying security questions remain the same:
“Who—or what—is trusted to act, what authority has been delegated, and how can that trust be continuously verified, constrained, and withdrawn when necessary?”
These are questions that PKI professionals have been answering for decades. Their experience may prove just as valuable in the age of AI as it has throughout the evolution of enterprise security.
Read the original article on LinkedIn ↗