What Evidence Should You Ask a Vendor for Before Giving It Access?

A vendor questionnaire is not enough. Learn what evidence to request before granting access, from permissions and MFA controls to security testing, monitoring, and revocation.

A vendor needs access to troubleshoot an application, manage infrastructure, or connect a service. The business wants the work started. The vendor says its security program is strong and sends a completed questionnaire. Is that enough to open the door?

Not by itself. A questionnaire describes what a vendor says it does. An access decision needs evidence of what will happen in your environment: which identities can connect, what they can reach, what actions they can perform, and how you can stop them.

The goal is not to collect the most documents. It is to obtain enough relevant, verifiable evidence to approve the smallest necessary access—or explain why access must wait.

Start With the Access Request, Not the Vendor’s Security Brochure

“Vendor access” can mean a named technician entering a support portal, a remote-management tool reaching servers, or an integration receiving permission to read and modify business data. Those arrangements create different risks and require different evidence.

Before reviewing security materials, define the proposed connection. Identify the business owner, systems involved, data exposed, requested actions, and access duration. Separate human accounts from service accounts and application permissions. An integration that runs without human interaction needs different authentication controls from a technician signing in remotely.

NIST’s security assessment methodology uses examination, interviews, and testing rather than relying solely on written assertions. Although that publication addresses a specific assessment context, the underlying discipline is useful here: review artifacts, clarify uncertainties, and verify important controls in proportion to the access risk.

A Short Checklist of Evidence to Request

Use these five requests as the starting point. Ask the vendor to identify each artifact’s date, scope, and exceptions.

  • Scope and necessity: A list of systems, data, named users, service accounts, requested permissions, business purposes, and access duration. Require a specific justification for administrative privileges.
  • Identity and connection controls: A configuration export, redacted screenshot, or demonstration showing MFA enforcement for human remote-access accounts, assigned roles, and the approved connection method. For integrations, request the actual application permissions or authorization scopes.
  • Testing and unresolved findings: A relevant independent assessment or vulnerability-testing summary, including its date, coverage, open findings, and remediation evidence. Confirm that the systems supporting your service are included.
  • Monitoring and revocation: A sample access log showing who connected, when, and to what; the person or team responsible for reviewing activity; and procedures for expiring accounts and revoking connections. Validate the revocation path before production access.
  • Incident responsibilities: A security contact, written notification commitments, and a clear division of investigation, containment, and recovery responsibilities between your organization and the vendor.

These requests align with the access-control practices in CISA’s Guide to Securing Remote Access Software and the supplier-management considerations in its guidance for operationalizing vendor supply chain risk management. Scale the depth of review to the potential impact, not simply the size of the contract.

Check Whether the Evidence Proves the Control You Need

Distinguish a policy from an enforced setting

A policy requiring MFA does not establish that every vendor technician’s account is covered. Ask to see enforcement for the relevant accounts and connection path, including exceptions. Likewise, a role called “support” does not establish least privilege. Review its actual permissions.

For machine-to-machine access, do not demand an interactive MFA screen that does not fit the authentication model. Instead, examine the application identity, credential protection, permission scope, and ability to rotate or revoke credentials.

Match assessment coverage to your exposure

An assessment of one product or business unit may say little about the service connecting to your environment. Check scope, exclusions, unresolved findings, and significant changes since testing. A report’s existence is not the same as assurance that the proposed access is safe.

Evaluate findings in context. A weakness affecting the remote-access path deserves different treatment from an issue in an unrelated environment. Ask what was fixed, how the fix was verified, and what exposure remains.

Accept secure alternatives, not unsupported assurances

Vendors may reasonably withhold detailed reports or sensitive configurations. Offer a secure review channel, controlled document access, redacted evidence, or a live demonstration. Do not request passwords, tokens, private keys, or customer records. However, confidentiality concerns do not eliminate the need to establish that essential controls work.

A Sample Evidence Request You Can Send

Subject: Evidence required before vendor access approval

Before we enable access, please provide:

  • The systems, data, identities, permissions, business purpose, and duration requested, including justification for administrative access.
  • Evidence of authentication enforcement, assigned roles, connection controls, and any application or integration permissions.
  • Relevant security-assessment or vulnerability-testing results, open findings, and remediation status.
  • A sample access log, the activity-review owner, and procedures for expiring access and revoking connections.
  • Your security contact, incident-notification commitments, and response and recovery responsibilities.

For each artifact, identify its date, the service or environment covered, and any exceptions. Redacted evidence or a secure demonstration is acceptable. Please do not send passwords, tokens, private keys, or customer data.

We will use this evidence to approve the defined scope, propose narrower access, or defer access pending remediation. Please identify anything you cannot provide and suggest an alternative way to validate the control.

How the Answers Change the Access Decision

The following is a practical decision rubric, not a set of approval categories prescribed by NIST or CISA. Each decision should follow from the evidence and the consequences of misuse.

Approve: The requested access is necessary and controlled

Approve when permissions match the business need, identity and connection controls are demonstrated, relevant findings are resolved or formally accepted by the appropriate risk owner, and monitoring and revocation are ready. Acceptance of residual risk should explain the exposure and safeguards, not merely record a signature.

Illustrative example: A support vendor requests read-only diagnostic access through your approved gateway. Its named accounts have enforced MFA, the role cannot change configurations, activity is attributable, and the access-expiration process has been tested. That evidence can support approval of the documented scope and duration—not broader or indefinite access.

Restrict: A narrower arrangement adequately contains the gap

Restrict access when the business task can proceed safely through a smaller permission set or a more controlled workflow. Options include read-only permissions, a test environment, supervised sessions, time-limited elevation, or customer-executed changes.

Illustrative example: A vendor asks for standing administrator access to diagnose an application problem but cannot justify continuous privileges. A supervised troubleshooting session with temporary elevation may meet the need without leaving an administrator account active afterward.

Restrictions must address the actual weakness. Shortening a session does not compensate for uncontrolled data export, and read-only access can still expose sensitive information. If the narrower arrangement leaves the material risk intact, delay instead.

Delay: Essential assurance is missing or the exposure remains unacceptable

Delay when evidence is missing, contradictory, or unrelated to the service; powerful permissions remain unjustified; or significant relevant weaknesses lack remediation or effective compensating controls.

Illustrative example: A vendor requests broad production access through a shared account, cannot demonstrate MFA enforcement, and cannot provide attributable activity logs. A general security policy does not close those gaps. Keep production access disabled and specify the controls and evidence required for reconsideration.

Record the Decision and Make It Expire

The approval record should connect the request, evidence, and decision. Document the business owner, approver, authorized identities and permissions, conditions, unresolved risks, expiration, and review date. State who can revoke access and how to reach that person during an incident.

Where your organization controls the gateway, identity platform, or application authorization, verify your own settings too. Vendor evidence cannot prove that your side of the connection is configured correctly.

Reassess access when the work changes, permissions expand, personnel change, or an incident affects the arrangement. Approval for troubleshooting should not quietly become approval for permanent administration.

Make Evidence the Gate to Access

A vendor’s reputation, certification, or completed questionnaire can inform your review. None replaces evidence tied to the connection you are about to authorize.

Before enabling the next vendor account, request the five evidence categories, verify the controls that matter most, and document an approve, restrict, or delay decision. If you cannot explain who has access, why they need it, how you will observe it, and how you will end it, the access request is not ready.

Browse all insights · Contact Bart McDonough