TL;DR: For Akeyless, the EU AI Act’s high-risk deadlines have shifted to December 2027 and August 2028, but AI agent access still raises unanswered questions around authorisation, traceability, and compliance for organisations also bound by GDPR, NIS2, and DORA. Delayed deadlines do not reduce the need to govern runtime permissions and audit evidence.
At a glance
What this is: NHI Mgmt Group's analysis, drawing on Akeyless's event briefing: The EU AI Act deadline change does not remove the need to govern AI agent identities, permissions, and auditability across regulated environments.
Why it matters: Identity and security teams still need defensible controls for AI agent authorisation, logging, and accountability because legal timelines shifted without changing runtime access risk.
By the numbers:
- The EU AI Act’s high-risk compliance deadlines have moved to December 2027 and August 2028.
👉 Read Akeyless's briefing on EU AI Act deadlines and AI agent identity security
Context
The core issue is not the date on the calendar but the governance model underneath it. AI agents can still request, combine, and use access across sensitive systems and data, so compliance teams must decide who authorises those actions, how they are bounded, and what evidence proves they were controlled.
For organisations operating in Europe, the delay to the EU AI Act does not remove overlap with AI agent compliance guidance under GDPR, NIS2, and DORA. The practical problem is proving that agent permissions, runtime decisions, and audit trails are aligned with regulatory expectations before the first material incident tests them.
Key questions
Q: How do you know if your identity governance model is keeping up with AI agents?
A: You know it is keeping up when it can explain each agent’s authority, delegation path, and review state in real time. If access is only checked at provisioning or quarterly review, the model is already behind. Strong governance for AI agents must account for runtime decisions, sub-agent creation, and fast revocation before the session closes.
Q: When should organisations use runtime authorization for AI agents?
A: Use runtime authorization when agent behavior can change based on context, tools, or delegated workflows. Static approvals are too coarse when an agent can act across multiple systems in minutes. Runtime checks help keep privilege proportional to the current task and reduce the chance that a one-time approval becomes persistent excess access.
Q: How can organisations make AI agent actions auditable?
A: Organisations need logs that connect each action to a specific agent identity, the delegator, the purpose, the tokens used, and the downstream systems touched. Auditability should cover the entire delegation chain, not just the final API call. If the record stops at the application layer, it will not support compliance, incident response, or accountability.
Q: What is the difference between access governance and audit governance for AI agents?
A: Access governance controls what an AI agent can reach before it acts. Audit governance records what the agent did, when it did it, and why. Both are necessary, but they serve different purposes. Access governance prevents harmful actions up front, while audit governance helps teams reconstruct events, investigate anomalies, and improve controls after the fact.
Background and context
Runtime authorisation for AI agents
Runtime authorisation is the decision point where an identity is allowed to take a specific action at execution time, not just at provisioning time. For AI agents, that matters because access is often task-specific, dynamic, and context-sensitive. A static role assigned once at onboarding cannot reliably express which tools, data sets, or actions an agent may use in a given session. The control problem is therefore less about assigning an identity and more about constraining what that identity can do when it acts. See the AI Agent Authorisation Guide.
Practical implication: Treat agent permissions as per-action decisions, not durable entitlements that can be reviewed only after deployment.
Audit trails for delegated AI actions
Auditability for AI agents depends on recording enough context to reconstruct who or what triggered an action, what data or tools were used, and whether the action stayed within policy. Traditional logs often capture system events but not the delegation chain or the policy decision behind each step. Without that context, incident response and compliance teams cannot distinguish legitimate autonomous activity from misuse, scope drift, or delegated overreach. See the AI Agent Observability, Audit and Incident Response Guide.
Practical implication: Capture action provenance, policy decisions, and session context so each agent step can be traced back to an accountable authority.
Identity governance under EU regulatory overlap
The EU AI Act now sits alongside GDPR, NIS2, and DORA for many European organisations, so AI agent identity governance cannot be treated as a single-law exercise. Each regime pressures a different part of the control stack: personal data handling, operational resilience, and accountable access. That overlap means agent identity, privilege, and logging controls must be defensible across multiple compliance lenses rather than tuned to one deadline or one policy domain. See the Agentic AI Compliance Guide.
Practical implication: Align AI agent governance evidence to the overlapping regulatory duties that govern the systems, data, and services the agents touch.
NHI Mgmt Group analysis
Deadline drift does not change the identity problem: the risk in regulated AI deployments comes from runtime access, not from the calendar date attached to the law. A delayed compliance milestone can create breathing room for programme planning, but it does not reduce the need to decide what an agent may do, when it may do it, and who can prove that decision later. Practitioners should treat the deadline shift as a scheduling change, not a control change.
AI agent governance is now a multi-regulation problem, not a single-Act problem: GDPR, NIS2, and DORA continue to shape the evidentiary burden even when one AI deadline moves. That means permissions, audit trails, and accountability must be designed to satisfy privacy, resilience, and access-control expectations at the same time. The implication is straightforward: identity teams need one control model that survives multiple regulatory views.
Runtime authorization is the control that matters when agent behaviour is not fully knowable in advance: role assignment at onboarding is too coarse for systems that can select actions in context. The control question becomes whether a specific action can be approved, bounded, and recorded at the moment it occurs. That shifts governance from static entitlement review to live decisioning, which is the only defensible model for sensitive agent access.
Auditability must capture the delegation chain, not just the system event: organisations need to know who authorised the agent, what policy allowed the action, and what context governed the step. Without that chain, compliance evidence becomes brittle and incident reconstruction becomes speculative. Delegation traceability gap: this is the named concept that matters here, because accountability breaks when the approving context and the executed action cannot be tied together. Practitioners should design for reconstructable authority, not just log volume.
From our research library:
- 69% of security leaders agree identity management must fundamentally shift to address agentic AI systems, according to the 2026 Infrastructure Identity Survey.
- Read next: AI Agent Authorisation Guide
What this signals
Delegation traceability will become a default expectation in regulated AI programmes: once AI agents are allowed to act on behalf of people or systems, regulators and auditors will expect reconstructable evidence for each material step. The organisations that prepare now will be able to answer who authorised the action, what policy permitted it, and which system state changed as a result.
Identity programmes that still treat agent permissions as static entitlements will struggle when access is granted, used, and revoked inside a single operating cycle. Governance has to move closer to the moment of action, because delayed review cannot explain a decision that already happened.
Only 44% of organisations have implemented any policies to manage their AI agents, despite 92% agreeing that governing AI agents is critical to enterprise security, according to the 2026 Infrastructure Identity Survey. That gap means most teams are still designing control intent faster than they are building operational evidence, which leaves runtime authorization and audit trails as the immediate priority.
For practitioners
- Define agent-specific authorization boundaries Map each AI agent to explicit actions, data scopes, and system boundaries before deployment so runtime decisions can be evaluated against a known policy envelope.
- Build evidence for delegated actions Require logs that preserve the authorising policy, session context, and execution outcome for every significant agent action, not just the final API call.
- Align controls to overlapping regulations Document how the same agent identity controls support GDPR, NIS2, and DORA obligations so compliance evidence is reusable across legal reviews.
- Review where human approval still belongs Identify the agent actions that remain too sensitive for automated execution and keep human approval gates where the risk or accountability burden demands it.
Key takeaways
- The main issue is not the postponed deadline but the unchanged need to control AI agent permissions at runtime.
- Compliance pressure now spans multiple regimes, so evidence for agent access must satisfy privacy, resilience, and governance expectations together.
- Strong agent governance depends on action-level authorization and reconstructable audit trails, not only on onboarding controls.
Related NHI Mgmt Group resources: Agentic AI Identity Guide · AI Agent Authorisation Guide · AI Agent Observability, Audit and Incident Response Guide · Agentic AI Compliance Guide
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 and OWASP Non-Human Identity Top 10 address the attack surface, NIST AI RMF sets the technical controls, and NIS2 defines the regulatory obligations.
| Framework | Control / Reference | Relevance |
|---|---|---|
| OWASP Agentic AI Top 10 | ASI03 — Identity & Privilege Abuse | Agent identity and privilege are the central governance problem in this briefing. |
| ASI09 — Human-Agent Trust Exploitation | The briefing centres on delegated authority and who can rely on an agent's actions. | |
| Recommendation — Constrain agent privileges to the minimum action scope and verify each runtime decision. Define where human approval remains mandatory before an agent can act on sensitive systems. | ||
| OWASP Non-Human Identity Top 10 | NHI-05 — Overprivileged NHI | AI agents are non-human identities, and the briefing focuses on excess permissions and runtime scope. |
| Recommendation — Review agent entitlements for overreach and remove privileges that are not required for the task. | ||
| NIST AI RMF | GOVERN — AI Governance and Accountability | The briefing is about organisational accountability for regulated AI agent use. |
| Recommendation — Assign clear accountability for agent decisions and document how oversight is enforced. | ||
| NIS2 | Operational resilience and security governance | NIS2 is directly relevant because the briefing ties AI agent controls to EU resilience obligations. |
| Recommendation — Align agent access and logging controls to the resilience requirements that apply to critical services. | ||
Key terms
- AI Agents: AI agents are autonomous software entities that act within organisational environments and make runtime decisions within assigned boundaries. They can hold identities, authenticate to systems, and exercise permissions, which makes them comparable to other non-human identities that require inventory, governance, and continuous activity monitoring.
- Runtime Authorisation: Runtime authorisation is the practice of deciding access while a task is in progress, rather than only at provisioning time. It matters for NHIs because credentials and entitlements can change risk mid-session, especially when automation or AI agents interact with sensitive systems.
- Audit Trail: An audit trail is a record of who accessed a system, what they did, and when they did it. For PHI environments, it provides the evidence needed to investigate incidents, support breach determinations, and demonstrate that access was attributable to a specific identity or workflow.
- Delegation Traceability: Delegation traceability is the ability to link an agent’s actions back to the originating user request, approval path, and tool use. It gives security and compliance teams evidence for investigations, policy enforcement, and post-incident review when an autonomous system accesses data or performs actions on behalf of someone else.
What to expect at the briefing
Akeyless's full event briefing covers the operational detail this post intentionally leaves for the source:
- How the revised EU AI Act timeline changes the compliance conversation for enterprise AI agents
- Practical discussion of where GDPR, NIS2, and DORA intersect with agent identity and access decisions
- Examples of least-privilege controls, runtime authorization patterns, and audit evidence expectations
- Audience Q&A with legal and strategy perspectives on AI agent governance
Deepen your knowledge
NHI governance, agentic AI identity, and machine identity lifecycle are core topics in our NHI Foundation Level course, the industry's only accredited NHI security programme. If you are building or maturing an IAM programme, it is worth exploring.
Published by the NHIMG editorial team on November 16, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org