Identity teams should prioritise sessions that translate emerging concepts into operating decisions. Focus on where machine identities are created, how their privileges are governed, and how agentic AI changes trust boundaries. The best learning agenda connects strategy with controls, so teams leave with practical guidance for governance, access review, and future policy design.
Why This Matters for Security Teams
Conference agendas often blur machine identity governance with general cloud security, but identity teams need a sharper filter. agentic ai introduces software that can decide, chain tools, and request access dynamically, which means the risk is not just “more secrets” but unexpected privilege use at runtime. Sessions grounded in operating decisions help teams understand where static IAM assumptions stop working and where controls must move toward policy evaluation, workload identity, and short-lived authorization.
The most useful talks connect these shifts to real breach patterns. NHIMG’s Ultimate Guide to NHIs notes that 97% of NHIs carry excessive privileges, which is a strong reminder that access sprawl is already the baseline before agents are introduced. Security teams should also watch for sessions that tie agent behaviour to the current guidance in the NIST AI Risk Management Framework and the OWASP Agentic AI Top 10, because those sources frame risk in terms of system behaviour, not just identity objects.
In practice, many security teams encounter agentic AI privilege failures only after a tool chain has already been abused, rather than through intentional governance design.
How It Works in Practice
Identity teams should prioritise conference content that shows how an agent gets its identity, how it proves it is allowed to act, and how access is revoked after the task ends. For machine identities, the most relevant material usually covers workload identity, short-lived tokens, and policy-as-code. For agentic systems, the key question is not “what role does this user have?” but “what context is this agent operating under right now?” That distinction is central to both CSA MAESTRO agentic AI threat modeling framework and MITRE ATLAS adversarial AI threat matrix.
Look for sessions that explain:
- how workload identities are issued and bound to a service, agent, or pipeline step;
- how JIT credentials reduce blast radius by issuing secrets only for a specific action window;
- how runtime policy checks decide whether a tool call is permitted in context;
- how audit logs show both the machine identity and the agent’s intent, not just the token used.
NHIMG’s OWASP NHI Top 10 discussion is useful here because it reinforces that agents can move laterally through tools when permissions are broad, stale, or inherited from human-centric IAM patterns. The most actionable conference sessions usually show how teams combine SPIFFE-style workload identity, OIDC-based attestation, and policy engines such as OPA or Cedar to enforce real-time decisions rather than pre-baked entitlements. These controls tend to break down in legacy environments where agents share service accounts, long-lived API keys, or unmanaged integration credentials across multiple tools.
Common Variations and Edge Cases
Tighter machine-identity control often increases operational overhead, requiring organisations to balance stronger containment against deployment speed and developer friction. That tradeoff is real, especially when teams are trying to support both traditional services and autonomous agents in the same platform. Best practice is evolving, and there is no universal standard for this yet, so conference sessions should be judged on whether they acknowledge those constraints rather than pretending one control pattern fits every stack.
Prioritise talks that address edge cases such as multi-agent workflows, delegated tool use, human-in-the-loop approval chains, and third-party integrations. These are the places where static RBAC breaks down fastest because an agent’s next action is not fully predictable in advance. The strongest learning sessions will explain how to scope privileges per task, how to separate orchestration identity from execution identity, and how to rotate or revoke credentials as soon as a workflow completes. NHIMG’s Ultimate Guide to NHIs is especially relevant when discussing why long-lived secrets and poor offboarding remain persistent failure points, while the NIST AI Risk Management Framework helps translate those lessons into governance language. If a session cannot explain how its guidance works when agents chain tools, inherit permissions, or cross trust boundaries, it is too abstract to guide conference learning priorities.
Standards & Framework Alignment
This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.
OWASP Agentic AI Top 10, CSA MAESTRO and OWASP Non-Human Identity Top 10 address the attack and risk surface, while NIST AI RMF and NIST CSF 2.0 set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| OWASP Agentic AI Top 10 | A2 | Agentic risk controls help teams judge sessions on runtime behavior, not static roles. |
| CSA MAESTRO | GOV-1 | MAESTRO stresses governance for autonomous workflows and machine identities. |
| NIST AI RMF | AI RMF frames accountability, monitoring, and risk management for AI systems. | |
| OWASP Non-Human Identity Top 10 | NHI-03 | NHI credential lifecycle is central to machine identity and agent access governance. |
| NIST CSF 2.0 | PR.AC-4 | Least privilege and access management underpin safe handling of NHIs and agents. |
Prioritise learning on short-lived secrets, rotation, and revocation for machine identities.
Related resources from NHI Mgmt Group
- How should aviation security teams reduce identity blind spots across human, non-human, and agentic AI accounts?
- How should security teams govern machine identity credentials in agentic AI environments?
- What do teams get wrong about machine identity in AI programmes?
- What do teams get wrong about machine identity security in AI programmes?
