Identity sprawl
Agents, tokens, models and workloads appear across cloud and AI platforms faster than ownership models can keep pace.
IDENTITY / CONTROL / EVIDENCE
CP / 2026 / ENTERPRISE
Compute Passport™ · AI Identity Infrastructure
Give every agent an identity.
AI identity infrastructure for the enterprise.
Create a persistent identity for AI agents and bind it to ownership, authority, approved models, workloads, policy, runtime status and evidence.
01 Unique ID
02 Tracking
03 Audit
04 Report
CPID: AGENT-019B
OWNEREnterprise Ops
MODELApproved · v4.2
WORKLOADAttested
POLICYBounded scope
EVENTIdentity verifiedCPID active · proof valid
THE ENTERPRISE REALITY01 / WHY NOW
AI agents are becoming operating identities.
They access systems, call tools, select models, delegate work and change state. The enterprise needs to know what is acting, who owns it, where its authority came from and whether that authority is still valid.
Authority ambiguity
On-behalf-of, delegated, autonomous and sub-agent execution create different authority paths and different lifecycle requirements.
Runtime change
A model, workload, region, tool or data boundary can change after commissioning. Identity must remain verifiable at the action point.
Evidence gaps
Access logs alone do not explain the full decision chain across identity, policy, approval, execution and outcome.
THE PLATFORM02 / ONE TRUST LAYER
Identity and telemetry in. Governed execution and evidence out.
Compute Passport™ complements the enterprise identity layer. It adds agent, model, workload, compute, policy and cryptographic context without replacing the systems already authenticating users or operating workloads.
01 / CPID
Unique ID
Assign each governed AI object a persistent, globally unique identifier linked to sponsor, purpose, issuer and current status.
Explore Unique ID →02 / LIVE CONTEXT
Tracking
Resolve delegation, model, workload, tool, destination and status across commissioning, operation, change and retirement.
Explore Tracking →03 / DECISION CHAIN
Audit
Preserve identity assertions, policy decisions, approvals, provenance, runtime context and outcomes as one evidence chain.
Explore Audit →04 / ASSURANCE
Report
Translate machine-speed events into audit-ready packages for security, risk, AI governance, procurement and executive review.
Explore Reporting →HOW IT WORKS03 / GOVERNED EXECUTION
A control path that follows the agent from identity to evidence.
The workflow binds persistent identity to short-lived proof, evaluates authority and context at runtime, enforces the decision where action occurs and preserves a reproducible record.
INPUTOwner · purpose · object
CONTROLCPID issuance + lifecycle
OUTPUTAccountable AI identity
ENTERPRISE INTEGRATION04 / COEXIST BY DESIGN
Connect the identity and control systems you already run.
Adapters normalize identity, model, workload, policy and evidence events while preserving each existing system of record and enforcement boundary.
IDENTITYEntra · Okta · Ping · PAM / IGA
AIFoundry · Bedrock · model registries
CLOUDAzure · AWS · GCP · Kubernetes
SECURITYSIEM · SOAR · gateways · DLP
GOVERNANCEGRC · audit · CMDB · ITSM
ENTERPRISE CONTROL MODEL05 / SHARED OWNERSHIP
Built for identity, security, AI, cloud and risk teams.
The control model connects sponsor accountability, engineering implementation, runtime review, revocation and evidence across organizational boundaries.
Identity & Security
Trust anchors, authentication integration, delegation, privileged control and revocation.
AI Platform
Model registry, gateway integration, agent registration and runtime telemetry.
Cloud & Infrastructure
Workload identity, region, network, keys, compute, attestation and containment.
Risk & Audit
Evidence requirements, exceptions, monitoring, review cadence and assurance.
ENTERPRISE USE CASES06 / CONTROL IN CONTEXT
Give every agent an identity before it reaches a material action.
Apply one trust model across enterprise workflows while preserving the policy, data, approval and evidence requirements of each operating domain.
Bound transaction and model-risk workflows.
Control transaction thresholds, model/data boundaries, approvals, destination limits and evidence.
Protect sensitive clinical and administrative workflows.
Purpose-limited access, minimum-necessary data, regional processing and qualified human authority.
Constrain privileged machine action.
Bind workload, environment, tools and emergency revocation to the agent operating in segmented systems.
Govern developer and operations agents.
Limit repositories, environments, secrets, model/tool use and production changes while preserving signed evidence.
Make automated service action attributable.
Attach purpose, role, record scope, permitted actions and escalation to every case workflow.
Trace automated supplier decisions.
Bind evidence source, supplier context, freshness, approval gates and re-review triggers.
COMPUTE PASSPORT RESEARCH07 / IDENTITY FOR THE AGENTIC ENTERPRISE
Research on the identity architecture autonomous work requires.
Technical perspectives on persistent agent identity, bounded authority, runtime enforcement, revocation and evidence.
From Runtime Action to Audit Evidence: Governing AI at Machine Speed
Enterprise AI governance cannot depend on reconstructing logs after the fact. Identity, policy, approval, runtime context and outcome need to become one reproducible decision record as the action occurs.
Read research →Persistent Identity, Ephemeral Proof: A Trust Model for Enterprise AI Agents
A durable identifier and short-lived credentials solve different problems. Separating them gives enterprises a stronger model for verification, least privilege, revocation and cross-platform agent identity.
Read research →The Agent Identity Gap: Why Enterprise IAM Must Extend Beyond Humans and Applications
AI agents introduce machine-speed authority, delegation and runtime context that traditional identity models do not fully represent. The enterprise needs a governed identity object for the actor performing the work.
Read research →
Build for enterprise AI