Cerbi / Application logging governance
Sensitive fields. Consistent policy. Evidence of what was governed.
Find unsafe logging, apply policy at selected application or OTLP log boundaries, and retain recorded outcomes. CerbiShield runs in your Azure tenant, beside your existing observability stack.
See it. Control it. Prove it.
Inside CerbiShield / Reporting
See policy outcomes, prioritize application risk, and follow the evidence retained for review.
Catch the risk. Explain the decision.
Prioritize the application. Review the reporting window. Follow the retained evidence. Four distinct dashboard views show the questions your team can investigate.
01 / Prioritize application risk
Move from a severity pattern to the application, rule, and field behind the finding.
CerbiShield / Governance dashboard
Review questionWhere should we investigate first?Severity and application context narrow the review before you open an individual finding.Inspect application risk02 / Explain governance outcomes
Separate passed, violating, redacted, blocked, relaxed, and error outcomes from policy finding counts.
CerbiShield / Governance dashboard
Review questionWhat did the governance boundary record?Use recorded outcomes to explain the result, then follow retained evidence into the review.Review governance outcomes03 / Follow retained evidence
Distinguish healthy summaries from retained detail and review the recorded governance outcomes.
CerbiShield / Governance dashboard
Review questionWhat evidence can we review?Follow the Reporting projection into retained records while keeping source quality and unavailable inventory explicit.Follow retained evidenceYour observability platform may already offer processing and redaction. Cerbi is relevant when a logging-control or evidence gap remains: inconsistent policy across teams, sensitive fields escaping controls, or no record of which policy governed a workload. Keep Splunk, Datadog, Azure Monitor, OpenTelemetry, and your existing destinations.
01One logging-governance lifecycle
For platform and security teams facing a customer-assurance or audit deadline: discover unsafe logging, define versioned policy, enforce it at supported boundaries, and prove what happened. Preserve useful debugging context where policy allows.
Discover → Define policy → Enforce → Prove
CerbiShield manages policy versions, rollout, observed coverage, and evidence over time. CerbiStream applies policy in selected applications; Gateway applies it on explicitly configured OTLP log paths. Available actions depend on the integration and policy.
Evaluate CerbiShieldOne governed event, one terminal outcome. Findings are counted separately. Relaxed activity remains traceable.
A policy decision, up close
Inspect the same fictional event before policy, with masking, and with an explicitly permitted relaxation.
Illustrative policy example · synthetic fields.
Not a product screen or live evaluation.
OutcomeRedacted
The configured masking rule protects the email field. Application, operation, and trace context remain useful for debugging.
02Investigation + evidence
Review findings alongside the application, environment, policy version, recorded outcome, and observation window where that context is available. Missing evidence cannot establish control coverage.
Application / Environment / Policy / Outcome / Evidence
See the application risk capture above, or open the complete investigation view in the dashboard tour.
03Reporting + audit
Bring policy outcomes, configuration changes, and authorization history into a reviewable evidence story.
Reporting summarizes observed governance activity. Audit records help explain who changed configuration and when. Deployments and Service Health show the state of the governance plane itself.
Evidence supports governance reviews and audit readiness. It is not certification or a guarantee of regulatory compliance.
Explore reporting and evidence04Governed Applications
Register the application. Authorize the workload. Make access revocable and the lifecycle auditable.
Create a stable, tenant-scoped governed application.
Associate existing Entra workload principals; assign viewer or editor access.
Validate the workload token and active installation entitlement for governed access.
Read back governance profiles and review the recorded lifecycle.
Revoke an assignment or disable the application; authorization fails closed.
Retain the authorization lifecycle, including changes and denials.
Microsoft Entra remains your identity authority. Cerbi validates real workload tokens and tenant-scoped assignments, backed by an active installation entitlement. Revocation and disablement invalidate authorization caches.
Explore workload authorization05AI logging / Pro++ option
AI-related logs can carry sensitive fields too. Apply the same logging-governance lifecycle to supported provider, model, agent, and tool metadata emitted through connected logging paths.
Systems, providers, models, dependencies, and agent/tool context where supported and emitted.
Data classifications, protected-field evidence, request/token metadata, and estimated cost where supplied.
Review recorded runtime outcomes. Raw prompts, responses, and secrets are not collected by default.
Visibility depends on instrumentation and emitted log metadata. This does not authorize agent actions, govern model behavior, or provide universal AI estate coverage.
Customer-hosted architecture
CerbiShield deploys as an Azure Managed Application into the customer’s tenant. Managed identities, server-side tenant isolation, and governed access support architectural control.
Customer tenant
Microsoft Entra
Managed Identity
Azure Managed Application
Customer-controlled infrastructure
Policy, deployment, service health, and evidence in the customer’s operating environment.Build-time signal + runtime evidence
Run the free Scanner independently. Review findings against your existing controls; if a recurring gap remains, evaluate CerbiShield on one workload with an agreed policy and evidence requirement.
A founder-led logging-risk review
Bring a Scanner finding, a sensitive-field concern, or an evidence requirement. Compare existing controls first. If a gap remains, agree one workload, one policy, an evaluation window, and evidence to review before deciding on a recurring CerbiShield deployment.
As AI systems take more actions across enterprise applications, teams need evidence of what happened and how policies were applied. Cerbi helps engineering teams govern application telemetry and review recorded policy outcomes, alongside their existing security and observability tools.
Industry context: OpenAI’s report on the Hugging Face incident and model misalignment.
