TL;DR: Runtime authority, identity and auditability become core controls in agentic AI operations, as Codenotary explains the shift from monitoring software to governing software that can act. Autonomous systems need tighter bounds and revocable access once they can write code, execute commands and take production actions.
Editorial analysis by NHI Mgmt Group, based on content published by Codenotary: “Codenotary Joins NVIDIA Inception to Accelerate Innovation in Agentic AI Security”.
Key questions
Q: Why do autonomous AI agents increase security risk compared with read-only assistants?
A: Because risk comes from authorised action, not just generated output.
Q: What are the signs that an AI application has too much authority?
A: A common sign is that a single model response can write to databases, send messages, start workflows, or spend resources without a separate check.
Q: How should security teams govern agentic AI as it moves into production?
A: Security teams should govern agentic AI as a class of non-human identity, not as a generic application feature.
Practitioner guidance
- Define agent identity before deployment Assign every autonomous agent a unique, managed identity so actions can be attributed, authorised and revoked across its lifecycle.
- Constrain agent permissions to the task Use task-scoped access and explicit policy checks so agents can only reach the systems and actions required for the current workflow.
- Log tool use and decision paths Capture each agent's tool calls, resource interactions and action sequence so security teams can reconstruct behaviour after an incident or anomaly.
Bottom line: Autonomous agents change the security problem from monitoring software to governing software that can act.
Glossary terms: Agentic AI, Agent Identity, Task-Scoped Access, Runtime authority
Explore further
View Full Forum → | NHI Foundation Course → | Our Services → | Read the full analysis →
Agentic AI security is becoming an identity problem, not just a model-safety problem. Once an agent can write code, execute commands or access infrastructure, the security question shifts from whether the model behaves well to whether its runtime authority is bounded. That makes identity, delegation and revocation central governance controls rather than back-end implementation details. Practitioners should treat every agent as a governed actor with explicit permissions, not as a feature of the application stack.
A few things that frame the scale:
- 53% of security leaders expect AI to run major portions of their infrastructure autonomously within the next three years, according to the 2026 Infrastructure Identity Survey.
- 69% of security leaders agree identity management must fundamentally shift to address agentic AI systems, according to the 2026 Infrastructure Identity Survey.
A question worth separating out:
Q: What is the difference between finding an AI agent and governing it?
A: Finding an agent tells you that it exists. Governing it requires ownership, approval status, credential visibility, access scope, and a review process that can remove or restrict it when conditions change. Discovery is the input; governance is the decision system that makes the discovery operational.
👉 Read our full editorial: Agentic AI security needs monitoring and control as autonomy scales