Security

Your data stays yours.

We built Custos AI for businesses that care about data security. That means EU servers, strong encryption, zero AI training on your data, and GDPR compliance built in from day one — not bolted on later.

AES-256 encryption at rest

All data stored in our database and file storage is encrypted using AES-256. API keys use AES-256-GCM and are decrypted only in memory at the moment of use — never written to logs.

TLS 1.2+ in transit

All data transmitted between your browser and our servers is protected by TLS 1.2 or higher. No unencrypted connections are accepted.

EU-hosted — Frankfurt + Amsterdam

All customer data is stored exclusively on EU servers: database and files in Frankfurt (Supabase), LLM proxy in Amsterdam (TransIP). No data is transferred outside the EEA.

Zero AI training on your data

Your chat messages, uploaded files and API usage are never used to train, fine-tune or improve any AI model — by Custos AI or any LLM provider we connect to.

Daily encrypted backups

Automated daily backups with 30-day retention and monthly restore tests. Target recovery time objective: 4 hours.

GDPR-compliant by design

Full Article 28 GDPR Data Processing Agreement available on all plans. Sub-processor list published and maintained. 48-hour breach notification commitment.

Infrastructure

Application (Next.js)Vercel EU edge network
Database + Auth + File storageSupabase (Postgres) — Frankfurt, Germany
LLM proxy + budget enforcementLiteLLM (Docker) on TransIP VPS — Amsterdam, Netherlands
Email deliveryLettermint (EU SMTP)
PaymentsStripe + Stripe Tax (EU)
AnalyticsPlausible Analytics (cookie-free, EU-hosted)

Sub-processors

SupabaseDatabase, auth, file storage — Frankfurt, DE
VercelFrontend hosting and edge compute — EU
TransIPVPS for LiteLLM proxy — Amsterdam, NL
LettermintTransactional email — EU
StripePayments and Stripe Tax — EU
Note: LLM providers (OpenAI, Anthropic, Google, Mistral) are not sub-processors of Custos AI. Under the BYOK model, customers connect their own API keys and maintain a direct relationship with each provider.

Report a vulnerability

If you discover a security vulnerability in our platform, please report it responsibly. We aim to respond within 48 hours and will credit researchers in our changelog.