CAM Protocol | AI Governance vs. CyberArk & SailPoint | SafeLocSkip to content
CAM PROTOCOL · ENTERPRISE API

The Governance Layer That Existing
Security Tools Don't Cover

CyberArk governs credentials. SailPoint governs identity. HashiCorp governs secrets. These tools address different layers from a release decision. AI agents or people request sensitive data or an action. At least M of N named human approvers must approve, and any single denial permanently vetoes the request. After the configured veto window, CAM makes one signed call to your system, delivers the result once to the approved destination, and discards it. CAM does not store the released data, and every step is recorded in a tamper-evident, hash-chained audit log.

Request details you submit, such as endpoint URLs and metadata, are kept with the audit record. Do not place sensitive data in them.

The Governance Gap by the Numbers

Why CAM Protocol Exists

$10.22M

Average US enterprise breach cost

IBM Cost of Data Breach Report 2025

13%

Of enterprises have adequate AI agent governance

Gartner 2025

$10.5T

Global cybercrime cost in 2025

Cybersecurity Ventures Official Cybercrime Report 2025

97%

Of AI-related breaches had no AI access controls

IBM Cost of Data Breach Report 2025

Where CAM Fits

Review the controls in your stack. CAM adds human approval for sensitive data access and actions.

CAM governs sensitive data access and actions requested by AI agents or people. At least M of N named human approvers must approve, any denial permanently vetoes the request, and the configured veto window must pass before CAM sends one signed call to the customer's system, which releases data or executes the action. CAM does not execute actions. CAM does not store the released data.

Security stack analysis
Perimeter & Network

Network Security

Palo Alto · Zscaler · Cloudflare

COVERED
Identity & Access

Identity & SSO

Okta · Microsoft Entra · Ping Identity

COVERED

Identity Governance

SailPoint · Saviynt · One Identity

COVERED
Privileged Access & Secrets

Privileged Access (PAM)

CyberArk · BeyondTrust · Delinea

COVERED

Secrets Management

HashiCorp Vault · AWS Secrets · Azure Key Vault

COVERED
Detection & Response

Endpoint / XDR

CrowdStrike · SentinelOne · Microsoft Defender

COVERED

SIEM & Detection

Splunk · Microsoft Sentinel · IBM QRadar

COVERED
Data & Cloud

Data Loss Prevention

Microsoft Purview · Forcepoint · Symantec DLP

COVERED

Cloud Security Posture

Wiz · Prisma Cloud · Orca Security

COVERED
AI Agent Governance

AI Agent Request Governance

No dedicated release approval workflow shown

UNPROTECTED
This review has not identified a human-approval workflow for sensitive data access or actions in the selected layers.
CAM Release Workflow

Human Approval for Sensitive Data and Actions

AI Agent Data and Action Requests

AI agents and people can request sensitive data access or actions.

Threshold approval

M-of-N Human Approval

Thresholds define how many named human approvers must approve.

Built-In Denial Veto

Any one denial permanently vetoes the request.

Requests from AI Agents and People

Requests can come from AI agents or people.

Signed Call to Customer System

CAM sends one signed call to the customer's system.

Tamper-Evident Audit Log

CAM records each decision in a tamper-evident, SHA-256 hash-chained audit log.

Competitive Analysis

CAM Protocol vs. Leading Security Platforms

Security platforms address credentials, identity, and secrets. CAM addresses sensitive data and action requests from AI agents and people.

CapabilityCAM ProtocolCyberArkSailPointBeyondTrustHashiCorpOkta
Primary PurposeSensitive data access and action requests with named human approvalPrivileged credential mgmtIdentity lifecycle mgmtEndpoint privilege controlSecrets managementIdentity & SSO
Release AuthorizationNamed human approval and denial vetoYes: one admin credential grants privileged accessYes: single workflow approver can grant accessYes: single approver controls privilege escalationYes: stolen token grants full secret access, no human gateYes: one compromised identity reaches all SSO-connected systems
AI Agent Data Access and Action RequestsRequests routed to named human approversNoneNoneNoneNoneNone
Named Human Approval ThresholdNamed human approval thresholdSingle approverSequential workflowSingle approverNone (token-based)None
Denial / Veto RightsAny denial permanently vetoes the requestNoneNoneNoneNoneNone
Configured Veto WindowWindow must pass before signed callSession-basedRole expirationSession-basedToken TTLSession-based
Signed Call to Customer SystemOne signed call after approval and veto windowNoNoNoNoNo
Governed Data Not Stored by CAMCustomer system handles governed releasesIdentity/credentials onlyIdentity/roles onlyEndpoint/credentials onlySecrets onlyIdentity only
Tamper-Evident Decision LogEach decision recorded in a tamper-evident, SHA-256 hash-chained logSession/credential logsRole change logsSession logsAudit logSystem log
Healthcare Data Release RequestsM-of-N human approval; any denial vetoes; window must passVia configurationVia configurationVia configurationVia configurationVia configuration

Competitive data is based on publicly available documentation.

Security Architecture

How CAM Is Built

Website data protection

The website is served over HTTPS/TLS. Encryption at rest for the contact database is managed by the infrastructure provider. See our trust page for the current data-handling details.

Purpose Limitation

CAM processes governance metadata, including request details and decision events.

Signed Call to Customer System

The customer's system handles each release or action.

See what is in place today and what remains on our roadmap.

Trust and data handling

Ready to Integrate CAM Into Your Stack?

Priority onboarding for healthcare, financial services, and AI infrastructure teams.