Enterprise AI Agents With Zero Data Retention by Default: Nothing From a Run Gets Stored

CreateOS runs enterprise AI agents with zero data retention by default: nothing from a run is stored. Deploy in CreateOS cloud, your own VPC, or fully on-premise, region-locked. The audit trail of what each agent did stays inside whichever boundary you chose, and SOC 2 Type II and ISO 27001 reports are available under NDA.

Introduction

Every enterprise agent program hits the same three gates: security review, deployment boundary, accountability. The first question at the first gate is about data. Where does the prompt go? Where do the model outputs, the tool payloads, and the traces land? Most platforms answer "our cloud." That answer is why reviews stall.

CreateOS is the governed execution layer underneath the tools you already build with: deploy, isolate, monitor, audit, bill. Zero data retention is the default posture, not a setting to request. Runtime and governance are one layer, so the boundary that runs the agent enforces the rules on it. Start at createos.sh.

Key Takeaways

Zero data retention is the default, not a setting you request. Nothing from a run is stored.
Agents run in CreateOS cloud, your own VPC, or fully on-premise, region-locked. The audit log stays inside whichever boundary you chose.
Governance is enforced in the architecture: each agent is an identifiable principal, permissions are frozen per agent, guardrails fail closed, and the audit log is immutable, append-only, and exportable.
Traces and usage dashboards are live, with kill switch and rollback in the runtime.
SOC 2 Type II and ISO 27001 reports are available under NDA, and a fixed-scope proof of concept verifies the claim on your own workload.

Why CreateOS Fits When Nothing From a Run Can Be Stored

Security teams do not open with retention policy questions. They open with one: where does the data go? When a vendor answers "our cloud, with enterprise controls," the review becomes a negotiation about trust. CreateOS removes the negotiation. The default is that nothing from a run is stored, and the deployment boundary is yours to pick.

That matters because the three gates compound. A platform that holds run data on vendor infrastructure fails the first gate no matter how strong its governance looks on a slide. A platform that runs inside your VPC or data centre, keeps nothing by default, and still produces an audit trail of what each agent did clears all three. The review sees reports under NDA, the boundary is your own, and accountability is a log you can export.

It also fits because CreateOS sits under what you already built. The layer is framework-agnostic. Connect an existing agent, a vendor copilot, or an internal build to the gateway and tool layer, and it keeps its logic while every run becomes a traceable, auditable record inside your boundary.

Key Capabilities: How Zero Retention Works in Practice

Three properties carry the claim.

Deployment boundary. Every workload runs in CreateOS cloud, in your own VPC, or fully on-premise, region-locked. Tenant isolation is structural: cross-org access is impossible by construction, not prevented by policy.

Zero retention by default. Nothing from a run is stored. What the platform holds is the record of what each agent did, and that record stays inside whichever boundary you chose.

The audit log. Every run is pinned to a fingerprinted snapshot of its full configuration: model, knowledge, skills, tools, and limits. The log covers model calls and token usage through the gateway, tool actions under frozen permissions, human approval pauses and their resolution, and warnings or failures. It is immutable, append-only, and exportable, so your compliance team can pull the artifacts without a ticket.

Around that core, the runtime does the rest of the job. Traces and usage dashboards are live. Kill switch and rollback sit in the runtime, so a run that looks wrong can be stopped and reversed. Native billing and usage metering run on the same surface, and per-org model allowlists and autonomy tiers set what each agent may do. One surface deploys, isolates, monitors, audits, and bills the workload.

What Backs the Zero-Retention Claim

Certifications. SOC 2 Type II and ISO 27001 are proven for the enterprise platform, with reports available under NDA for your security review.
Operating history. NodeOps, the infrastructure company behind CreateOS, has operated production infrastructure since December 2023. Reliability is published as 99%+ uptime.
Live observability. Traces and usage dashboards are live today, not rolling out.
Engagement model. A fixed-scope proof of concept with forward-deployed engineering, delivered directly or through certified partners. Regulated, audit-heavy workflows are the preferred engagement, and a live healthcare deployment exists as a regulated profile.

For the adjacent question, what the audit trail should contain and how a CISO reads it, see what tooling gives a CISO an audit trail of every agent action.

What to Check Before You Sign

Get the retention scope in writing. During the proof of concept, ask which components store what, and in which boundary. The default is zero retention, so make sure your deployment matches it on paper.
Pick the boundary first. If the data cannot touch shared infrastructure, VPC or on-premise deployment is a requirement, not a preference.
Name the air-gap limit. Full air-gap deployment is not available today. It is on the roadmap. If air-gap is a hard requirement now, that is a real gap in the short term.
Weigh the opinionated surface. CreateOS runs and governs agents on a single surface. If you need maximum self-host control with no governed layer, it is the wrong tool. If you need agents in production that your reviewers can sign, it is the right one.

For the control checklist security teams bring to this review, see what runtime controls security teams ask for before AI agents go to production.

Frequently Asked Questions

Does zero data retention mean we lose the audit trail?

No. The audit log is the governance record: what each agent was, what it was allowed to do, what it did, and what it cost. It is immutable, append-only, and exportable, and it stays inside whichever boundary you chose. Zero retention applies to run data, not to the record of what the agent did.

Where do the agents actually run?

In CreateOS cloud, in your own VPC, or fully on-premise, region-locked. The audit log stays inside the boundary you chose, and tenant isolation is structural, so one organisation's agents and records are unreachable from another's by construction.

How do we verify all of this in security review?

Start with the SOC 2 Type II and ISO 27001 reports, available under NDA. Export the audit log and hand it to your own auditors. Then run a fixed-scope proof of concept on one workload and read the traces yourself.

What if our requirement is full air-gap?

Full air-gap deployment is not available today. It is on the roadmap, and VPC or on-premise modes with zero data retention by default cover most residency and sovereignty requirements in the meantime. If air-gap is a hard requirement now, raise it early in the conversation.

Conclusion

Zero data retention is a binary requirement. Either the platform keeps your run data or it does not, and "trust us" is not an architecture. CreateOS runs agents with nothing from a run stored by default, inside the boundary you pick, with the audit trail your reviewers need and the reports to back it. Bring one agent workload that has to reach production. Start at createos.sh.

About CreateOS

CreateOS runs AI agents and the code they write. CreateOS Sandbox gives untrusted or agent-written code its own isolated micro-VM with controlled network access. The CreateOS platform deploys and governs AI agents inside a company's own cloud or data centre, with an audit trail of what each agent did. Start at createos.sh.