AAM Self-assessment

Agentic Access Management

A Comprehensive Approach to AI Identity and Access Security

AAM Self-assessment

Benchmark your organization’s readiness to govern AI-agent access. In 10 quick questions, select every statement that applies; if none apply, leave the item blank. You’ll receive an overall maturity score and prioritized actions to guide investment and accountability. (~3–5 minutes)

1 Do you maintain a comprehensive inventory of AI-related Non-Human Identities (NHIs)?

Select all that apply:

AI identities are cataloged

Inbound access to AI identities is documented.

Outbound access from AI identities is documented.

Authentication methods for each identity are documented.

2 Do you automatically detect and remediate unauthorized or newly introduced AI services?

Select all that apply:

AI-related websites are detected and governed.

AI software running on endpoints is detected and governed.

Model providers and AI agent frameworks used by services are discovered.

We have enforceable allow/deny policies for AI tools and services.

3 Have you assigned clear human ownership and lifecycle accountability for every AI service and its identities?

Select all that apply:

Every AI identity has an assigned human owner.

Unused or orphaned accounts are periodically detected and handled.

Contextual workflows and activity signals trigger automatic ownership revalidation or decommissioning.

4 Do you provision AI-related identities only through approved workflows?

Select all that apply:

There is a dedicated process for AI identities.

An owner is required before activation.

Permissions are explicitly scoped (least privilege).

Business justification is documented.

Credentials are shared and stored securely.

5 Credential Hygiene & Separation

Select all that apply:

Human-issued identities/credentials are never used by AI.

Short-lived credentials are used whenever possible.

When long-lived credentials are necessary, they are securely shared and stored.

When long-lived credentials are necessary, they are rotated frequently.

6 Do you enforce resource-consumption controls for AI systems?

Select all that apply:

Models have token quota limits.

Model token quota limits are set appropriately for our use cases.

API and tool usage by AI has rate limits.

Database usage by AI has rate limits.

7 Do you assess and control AI agents that have access to both sensitive data and the public internet?

Select all that apply:

We maintain a list of AI agents with access to sensitive data.

We maintain a list of AI agents with access to the public internet.

We conduct risk assessments for agents that hold both types of access simultaneously.

Public internet access is limited to a small set of approved services.

8 Do you classify and assess AI vendors and service providers for security, reputation, and data sovereignty?

Select all that apply:

We track vendors and SaaS solutions that use AI to access our organizational data.

For every model we use, we track the owning vendor and hosting location.

We perform regular trust assessments of those vendors.

We use vendor trust assessments to prioritize and remediate access policies.

Models accessing organizational data run only in pre-approved regions aligned to residency/sovereignty requirements.

9 Monitoring & Threat Detection Coverage

Select all that apply:

Inbound access to AI identities is logged.

Outbound access from AI identities is logged.

AI identity audit operations are logged.

Logs are tamper-resistant.

Baselines are learned and access anomalies are flagged.

Resource consumption is monitored and cost-attributed.

10 Risk Management & Continuous Improvement

Select all that apply:

We have defined KPIs that measure our maturity in securing AI access.

Our KPIs also measure how well we enable AI for the organization.

We have measured ourselves against those KPIs.

We regularly assess maturity, identify gaps, and drive improvements.

AAM Self-assessment Results

Here’s your leadership snapshot of AI access governance. Your score highlights strengths and gaps across the core pillars, with recommended quick wins and strategic next steps, so you can sequence investments, assign owners, and accelerate secure AI adoption.

1. Discovery & Inventory

1.1

Maintain Comprehensive Inventory of AI-Related NHIs

Establish and maintain a comprehensive inventory of all AI-related non-human identities (NHIs) - for example, roles, service accounts, service principals, and database users. The inventory must document which agents and systems each NHI corresponds to.The inventory must capture both inbound access used to interact with agents, and outbound NHIs used by agents to access external tools, resources, and data.

1.2

Document NHI Authentication Methods

For each NHI in the inventory, document the authentication methods employed, including authentication type (user access, delegated access, machine-to-machine access), secret storage location and security - when static credentials are used (e.g., secure key vault vs. hardcoded credentials) - OAuth configuration details including discovery endpoints where applicable, and principals who granted consent for delegated access scenarios.

1.3

Map NHI Consumer Relationships

For each NHI recorded in the inventory, identify and document its active usage status and potential, authorized, or known consumers (such as users, systems, or applications). Refer to Section 6.1 for requirements related to analyzing access patterns and frequency of use as part of ongoing monitoring.

1.4

Inventory AI System Resource Access

Document all resources accessible by AI-related NHIs, including but not limited to: (1) Database access (hosting location, sensitive data presence, access controls granularity, and read/write permissions); (2) SaaS integrations (access levels, service type, trust level and reputation); and (3) Internet access (general vs. restricted access and approved external destinations). See Section 4.2 for management of Data Exfilteration risks.

1.5

Detect and Manage Shadow AI Usage

Continuously identify unauthorized or unmanaged AI tools and services across environments using identity provider signals and endpoint telemetry. Implement automated detection for unapproved AI tools and establish remediation workflows for discovered shadow AI usage.

2. Ownership & Accountability

2.1

Assign Ownership to All AI NHIs

Ensure every AI-related NHI is assigned to a responsible owner (individual or team) in accordance with organizational governance requirements. Ownership may be assigned directly to human owners or to CMDB items with designated human owners. For instance, each AI agent may have a clearly defined primary owner, and each associated identity may include an additional owner where appropriate, based on the resource type and scope.

2.2

List humans accountable for AI NHI changes

Track and retain logs of all humans who manage, or provision static credentials (such as tokens, API keys, certificates) used by AI NHIs or who modify AI NHI configurations and permissions.

2.3

Detect and Decommission Unused or Orphaned Identities

Establish processes for ongoing NHI management including regular access reviews and revalidation, automated detection of unused or orphaned identities, and periodic ownership verification. Enable automated decommissioning of unused identities based on contextual signals and customized workflows. See control 3.1 for initial provisioning security requirements and workflow customizations.

2.4

Revoke or Reassign Access During Offboarding and Personnel Changes

Upon employee offboarding or role changes, identify and remediate all AI-related access, including NHI ownership reassignment, credential revocation or rotation, and documentation of access transfer to new owners.

3. Credential Lifecycle & Hygiene

3.1

Enforce Secure Agent and NHI Provisioning

Require AI-related NHIs to be created and initially provisioned only through approved workflows that ensure scoped access based on least privilege principles, vault-backed credential storage, assigned ownership before activation, and documented business justification for the NHI request. Implement automated policy compliance checks during provisioning and enable workflow customization based on geographic, departmental, or risk profile requirements.

3.2

Prioritize Short-Lived Credentials

Use ephemeral or system-managed credentials where possible to reduce credential leakage risk and simplify lifecycle management.

3.3

Implement Secure Secret Storage

Store all AI NHI secrets and credentials in approved secure key vaults or secret management systems. Prohibit hardcoded credentials in configuration files or source code.

3.4

Rotate Long-Lived Credentials

When long-lived credentials are necessary, enforce regular rotation according to enterprise security policies and support automated rotation where technically feasible. Implement automated detection of credential rotation policy violations and establish escalation procedures for non-compliant credentials.

3.5

Prevent Human Credential Reuse

Detect and flag instances where human-issued credentials (session tokens, browser cookies) are inappropriately used by AI systems or automated processes.

4. Access Security & Risk Management

4.1

Implement Resource Consumption Controls

For AI NHIs with access to consumable resources, implement usage controls including rate limiting mechanisms, quota enforcement, cost management boundaries, and LLM access limitations.

4.2

Manage Data Exfiltration Risks

For AI NHIs with sensitive data access and outbound connectivity, assess data exfiltration risks, evaluate potential for external influence through adversarial inputs, and implement data loss prevention controls appropriate for AI systems.

5. Vendor & Service Trust Management

5.1

Classify AI Services and Vendors

Maintain a comprehensive catalog of all AI services and SaaS platforms that use NHIs to access organizational resources and data. Tag each AI service with comprehensive metadata including model vendor, hosting vendor, deployment model (SaaS, self-hosted, cloud-managed), service type classification, and vendor trust assessment.

5.2

Maintain Service Reputation Framework

Assign and regularly update reputation scores for AI services and SaaS platforms based on data handling practices, security incident history, geographic and regulatory compliance, and use reputation scores to prioritize and remediate access policies violations.

5.3

‍Enforce Geographic and Sovereignty Controls

Ensure AI services and LLMs operate only in pre-approved cloud regions that align with data residency and sovereignty requirements.

6. Monitoring & Threat Detection

6.1

Monitor AI Authentication Patterns

Log and analyze inbound and outbound access events, including authentication methods and frequency, consumer access patterns, unusual or anomalous authentication behavior, and cross-system access correlations. This ongoing monitoring complements the baseline consumer mapping established in control 1.3.

6.2

Track AI Resource Consumption

Monitor and log AI NHI outbound resource usage patterns including API call volumes and patterns, data access behaviors, resource consumption trends, and cost attribution and tracking.

6.3

Detect Behavioral Anomalies and Policy Violations

Implement automated detection of unusual AI NHI behaviors including access pattern deviations, unexpected resource consumption, suspicious data access or transfer activities, overprivileged AI NHI accounts, and violations of data access boundaries. Cross-reference behavior with known threat indicators and established organizational policies.

6.4

Maintain Immutable Audit Logs

Log all AI NHI-related activities in tamper-resistant audit logs, including access events, policy decisions, and lifecycle operations, with retention per organizational policy.

7. Risk Management & Continuous Improvement

7.1

Implement Risk-Based AI Access Management

Prioritize AI NHI security controls and management actions based on risk assessment considering access patterns and privileges, data sensitivity exposure, policy compliance status, and behavioral anomaly indicators.

7.2

Track Governance Maturity Metrics

Monitor organizational AI NHI security maturity through KPIs such as percentage of AI NHIs with assigned ownership, credential rotation compliance rates, policy coverage and exception rates, mean time to detect and remediate violations, and innovation enablement metrics (balance between security and operational flexibility).

7.3

Support Continuous Improvement and Remediation

Establish feedback mechanisms to assess policy effectiveness, identify gaps in coverage or control, measure impact on business operations and innovation, and drive iterative framework improvements. Support automated or semi-automated remediation actions for policy violations, including credential revocation, rotation, and ownership reassignment workflows.

Retake Assessment Print/Save Checklist as PDF

Oops! Something went wrong while submitting the form.

The Agentic Access Management Framework

The Agentic Access Management is a cross-industry collaboration
This framework is provided for informational purposes only. Implementation is at your own risk, and neither Sequoia nor Oasis accept any liability arising from the use of this framework.