Privacy & data minimization
Your data stays yours — and so does your customers'.
Lizrd finds and fixes cloud waste by reading your infrastructure, not your data. Most teams need nothing beyond our defaults. For organizations with stricter needs — especially those that host infrastructure for their own customers — a few opt-in controls go further.
Every organization
On by default
✓ We never read your workload data
No database or bucket contents, no Kubernetes secrets, no environment values, no logs. Lizrd reads cost and utilization metrics and read-only control-plane APIs — nothing inside your systems.
✓ Infrastructure metadata, minimized
Identifying tag values (Owner, Customer, Project, emails) are one-way hashed before they are ever stored.
✓ Names never reach our AI provider
Resource names, repository names, and file paths are tokenized out of every AI prompt and restored only for your own view — the model sees placeholders, you see the real thing.
✓ Scrubbed, encrypted, read-only
Provider errors are scrubbed of secrets before storage, credentials are encrypted at rest with AES-256-GCM, and Lizrd never changes your infrastructure — every fix is a proposal you apply.
Opt-in · administrators only
Privacy Mode
Two controls for privacy-sensitive organizations. Each is a deliberate trade: more privacy for a little less precision. Leave them off and the product works at full strength.
Control 1
Don't ingest source code
Enables Your code is never stored or fetched. Lizrd grounds everything on infrastructure facts only.
Trade-off Step-by-step "Guide me" fixes become prose-only — you keep the exact recommendation, but not the repo-linked code diff.
Control 2
Don't use AI correlation
Enables Resource names aren't sent to our AI for name-based matching of resources to code.
Trade-off Slightly less automatic linking of resources to the code that owns them. Rule-based matching still runs.
Turning Privacy Mode on also cleans up the past. The moment you disable a control, Lizrd permanently purges the data it already held under the old setting — source snippets and source-grounded AI content when you turn off source ingestion, and the AI-matched resource→code links when you turn off AI correlation. You'll see a summary of exactly what was removed.
At a glance
| Setting | What it protects | What you give up |
|---|---|---|
| Defaults (everyone) | Workload data, secrets, tag values, and names sent to the AI | Nothing |
| + Source off | Your source code is never stored or fetched | Exact code diffs in fix guidance |
| + AI correlation off | Resource names aren't sent to the AI for matching | Some automatic resource → code linking |
Which should you use?
- → Most teams: leave the defaults. They're already privacy-minimizing, and you keep every feature at full strength.
- → You host infrastructure for your own customers, or have strict contractual requirements: turn on "Don't ingest source code."
- → Your resource names themselves are sensitive (e.g. named after your end-customers): also turn on "Don't use AI correlation."
An administrator turns these on in Settings → Privacy; they take effect immediately. See how we keep you safe overall on the Security & Trust page.
Optimize without oversharing
Connect read-only, keep your defaults — or dial privacy up. Either way, nothing changes until you say so.