What Peeve reads, stores, and never sees.

Plain statements of what Peeve does and doesn't touch, written so you can drop them straight into a security review.

DataStoredRetention
Element fingerprints, route mapYesUntil you delete
Session step tracesYesYour setting · 0–90 days
Your users' passwords or session tokensNever–
Customer data used for model trainingNever–
What about prompt injection?

Page text enters the model as data, with instruction and content strictly separated. The model requests capabilities by name. It cannot compose a request. An injection suite runs in CI on every deploy.

Certifications
SOC 2: on the 2027 roadmap
GDPR: sub-processor list published
Signed DPA: Scale plan
BAA: not offered today
Every status has a word beside it, never a colour alone. Nothing is claimed early.
Sub-processors are a real table with names, purposes and regions: see the trust center.
Safe by construction

Three boundaries the design won't let Peeve cross.

01
Read-only until you widen it

Peeve starts with no ability to act. You grant capabilities one at a time in peeve.config.json, reviewed in a pull request. Nothing runs that you didn't add.

02
The model never holds a credential

Capabilities point at your own endpoints; your execution service makes the call with your keys. The model requests a capability by name. It never sees a token or composes a raw request.

03
Every action states its effect first

Before anything writes, the effect is shown and (outside read-only scopes) confirmed. A single command is the kill switch: it disables all acting at once.

The credential boundary

The model asks by name. Your keys never cross the line.

The model requests a capability by name: change_plan, not a URL and not a token. Only that name crosses into your infrastructure, where your execution service holds the credentials and makes the call. An injection can ask for a capability; it can never compose a raw request or reach a key.

TRUST BOUNDARYyour infrastructure →change_planModelholds no credentialExecution serviceyour keys stay hereYour API
Questions

Is AI customer support safe for customer data?

It depends on the design. Peeve is read-only by default, and the model never sees your users' passwords or session tokens; it requests capabilities by name. Each customer's memory is encrypted at rest, isolated to your workspace and never used to train outside AI models. A sub-processor list is published and a signed DPA comes with the Scale plan.

14-day trial, no cardInstalls in 5 minutesRead-only until you widen itKill switch in one commandCancel anytime. No ONE question asked.
The one question we'll ask on the way out: which answer did Peeve get wrong? It is how the Brain learns. Conversations where Peeve found nothing are never billed.

Read-only until you widen it. 14 days free. No card. No call.