Skip to content

Data Awareness (Preview)

Data Awareness shows what kinds of data move through your LyftData deployments and where that data goes. It reports when sensitive data crosses a boundary and keeps the evidence behind each finding available for inspection. The page describes its job as: Ask your estate, then reveal sensitive movement, reconciliation, evidence, and review work.

Open Data Awareness at /awareness. The page includes these sections:

  • Sensitive flows: known business, sensitive, and technical data types, plus destinations and data movement. For a selected destination, you can see the source flows behind it.
  • Reconciliation: shows how a confirmed classification leads to a rule-derived finding and then to a governed outcome.
  • Estate review queue: review items for the estate, including the configured crossing that needs the most attention.
  • Evidence: evidence is shown on demand in a chain-of-custody drawer rather than inline.
  • View related telemetry: opens Observability for the related Job and Workflow activity. Data Awareness remains the authoritative governance view. It does not treat datasets or findings as telemetry signals.

When Data Awareness is enabled, a separate panel on the page shows the background AI analysis work.

Data Awareness requires the data-awareness license feature. That feature depends on ai-runtime, the embedded AI runtime, and takes effect only when both are granted. Current Community+, Standard, and Enterprise offers include them; Community Edition does not. Your signed license is authoritative. See Licensing.

A license is necessary but not sufficient. Background analysis also needs a configured provider account and recorded consent.

  1. Configure a provider account. In Settings → AI integration (/settings/integrations/ai), a platform admin adds a provider account. You can use an API key or sign in with ChatGPT. Then choose a model and verify the connection. Provider usage may incur charges with that provider.
  2. Enable Data Awareness. Setting up a connection does not turn on either AI feature. In Independent AI controls, an admin chooses Enable Data Awareness…, reviews the dialog, and selects Enable background analysis. Consent applies to one tenant and environment and to the provider accounts listed in the dialog.

By enabling it, you choose continuing background provider analysis. That analysis also runs when nobody is using the Operator, and you do not approve each analysis. Data Awareness consent does not turn on the LyftData Operator, and it does not grant separate write or effect approvals.

The consent dialog asks you to confirm that the following can leave this installation and go to the listed provider accounts:

  • Data Awareness analysis instructions
  • Tenant, resource, and subject identifiers
  • Scoped evidence summaries and references
  • Dispatch metadata
  • Results of permitted Data Awareness reads

The dialog says this is not a guarantee of redaction or metadata-only transmission.

To withdraw consent, open Settings → AI integration → Independent AI controls and choose Disable Data Awareness. Disabling Data Awareness retires its queued work and signals known in-flight runs to cancel. It does not change the Operator.

  • Cancelling provider work that is already running is best effort, so some outcomes may remain uncertain.
  • Data that has already been transmitted cannot be recalled.
  • If the control shows Off · cleanup pending, no new work is authorized until cleanup finishes. Use Retry withdrawal to settle the remaining local work.
  • The Data Awareness agent is read-only. It runs on the embedded runtime with the canonical Data Awareness Pack and a fixed set of eight read-only capabilities. Before each call, the runtime checks the Pack, capability version, arguments, and tenant again. It rejects raw samples returned by a capability.
  • Raw samples are not disclosed by default. Evidence views report a Raw sample posture for each item, so you can see whether any raw samples were returned.
  • People still decide. You confirm a classification from a review item with Confirm classification, and rule-derived findings follow from what you confirmed. Operational changes belong to the LyftData Operator and its approval flow, not to Data Awareness.
  • Data Awareness is Preview in 3.0.
  • Data Awareness and the Operator have separate consent. Turning one on or off does not affect the other.
  • Cancelling provider work that is already running is best effort, and data that has already been transmitted cannot be recalled.