Every action, every prompt, every inference in the audit trail.
Limit Platform records every action across every layer in an evidence trail the auditor, the CISO, and the regulator can use. Audit reviews reconstruct the platform state, not just the outcome.
// governance · in the audit trail · inside the perimeter
The audit trail is the platform, not a feature on top.
In a regulated AI deployment, "what happened" is not enough. The auditor needs to reconstruct why a decision was made, what the system saw at the time, which model version was active, who was authorized to act, and what policy applied. That reconstruction has to be possible six months, six years, or a regulatory cycle later.
Limit Trace is the platform layer that writes this evidence at every layer. Identity events. Authorization decisions. Storage reads and writes. Messaging activity. Compute job lifecycle. Model inference, with the input the model saw and the output it produced. Customer-system integration events. All retained, all queryable, all exportable in formats the audit needs.
Reproducible reviews, not narrative summaries.
- 01
Comprehensive event capture across every layer
Every platform action lands in the evidence store with timestamp, actor, target, policy applied, and outcome. No layer is exempt. No event is sampled away.
- 02
Inference-level evidence for AI decisions
Every model invocation records the input the model saw, the model version, the inference parameters, the output, and the downstream action taken. Reviews can replay the inference deterministically when the model is reproducible.
- 03
Policy-versioned, retention-controlled storage
Retention policies are configurable per event class to match the customer's legal and regulatory obligations. Policies are themselves versioned in the evidence store.
- 04
Query and export for audit workflows
Auditors query the evidence store directly through scoped, authorized access. Exports produce per-case packages in JSON, CSV, and PDF, structured for the audit team's existing review process.
Evidence stays where the work happens.
The evidence store deploys inside the customer perimeter alongside the workload it observes. No platform action ships its evidence elsewhere. Cross-site aggregation, where used, is customer-controlled and customer-bounded.
The store is append-only by design. Earlier records cannot be silently revised. Lifecycle actions (retention expiry, legal-hold release) are themselves audited.
- Where it lives
- In the same perimeter as the workload. Append-only by design. Sized per event volume and retention horizon.
- Who can query it
- Authorized auditors, scoped per role and per matter. Every query is itself logged.
- How long it lasts
- Per-event-class retention configured by the customer to match legal and regulatory horizons. Legal holds suspend expiry on demand.
- How it exports
- Scoped exports to JSON, CSV, and PDF for the audit workflow. Cryptographic integrity verification on every export.
This capability is one layer of the operating system. Every application uses it. Every customer perimeter that runs Limit Platform gets it. See the full architecture →
See it in your environment.
Walk us through the systems already in place. We'll show you how this layer fits.