Verax

Trust

Verax, profiled by Verax

We ask you to send us log exports and identity data. It would be incoherent to audit how AI vendors handle data and not publish our own answers — in the same format, to the same standard, as every vendor profile we write.

Data-handling profile

Trains on your input
No.Your ingested logs and findings are never used to train any model, ours or anyone else's, and are never added to the public vendor dataset.
What we collect
Only what you send: log events (timestamp, destination hostname, source IP, username, byte counts) and, if you use the Entra importer, OAuth grant metadata. We never receive request or response content — no prompts, no files, no traffic decryption.
Watch subscriptions
Email address and the vendors you watch. Never shown publicly, never shared or sold. One-click unsubscribe in every alert.
Data region
EU. Application in Paris (cdg), database in the EU (eu-west-3).
Retention
Assessment data lives in your workspace until you replace it or ask for deletion; each ingest replaces the previous run. Deletion on request, no retention floor.
Isolation
Every client has a private workspace keyed to their API key. Findings are scoped per workspace; there is no cross-tenant read path.
Sub-processors
Vercel (application hosting), Supabase (EU Postgres), Brevo (alert email delivery), Google Gemini (classifies changes in public vendor documents — never your data).
Your data and the LLM
No customer data is ever sent to any language model. The only text an LLM sees is publicly published vendor policy documents we fetched ourselves.
Deployment options
EU-hosted SaaS today. Self-hosted ingestion for organisations that will not send logs off-premises — ask us.
Exportability
Your findings are yours: every assessment surface is available over the API (JSON, OCSF, CEF) and as a report you can export at any time. No lock-in, no export fee.
Status
Independent product, sole operator. No SOC 2 or ISO 27001 audit yet — stated plainly rather than implied. If a certification is a requirement for you, tell us and we will say honestly where we are.

Why this page exists

Our methodology says unverifiable facts are marked unknown rather than guessed. The same rule applies to us: everything above is what we actually do, including the parts that are not flattering.