Security & Trust

Your data runs through our employees. Here is how we protect it.

We would rather earn your trust with clear commitments than with marketing language. Below is how we handle your data, the standards we build to, and how to reach our security team.

How we handle your data

Six commitments we hold for every customer, on every task.

We never train on your data

Your proprietary business data is processed only to complete your task. As a matter of policy, it is not used to train models, and it is not retained beyond what the task requires.

Encrypted in transit; credentials sealed at rest

All traffic is TLS-encrypted on the wire, with HSTS enforced. Your OAuth tokens, API keys, and integration credentials are sealed at rest with per-user envelope encryption.

Least-privilege access

Every employee acts under a scoped identity with read-only access by default. Write actions require explicit permission. There is no master key.

Every action is logged

A timestamped audit trail records what each employee did and why, so you can trace any decision after the fact — not just trust that it happened.

Export or delete, anytime

Your data is yours. You can export it, and you can request complete deletion, in line with your rights under GDPR and PIPEDA.

You stay in control

High-risk actions wait for your approval rather than running on assumption, and a kill switch stops any employee instantly. Oversight is the default, not an add-on.

Built to recognized standards

We build to established security and privacy standards rather than inventing our own.

NIST

Identity, authentication, and password-storage guidance (SP 800-63B).

OWASP

Application and LLM security practices (ASVS and the OWASP Top 10 for LLM Applications).

GDPR & PIPEDA

Data-protection rights under EU and Canadian law, including access, portability and erasure.

SOC 2 Type II is on our roadmap. We will begin a formal observation window after launch. Until that report exists, we will not claim a certification we do not hold.

What a security reviewer will ask

Plain answers, including the ones that are not finished yet.

Hosting & region
Backend and Redis on Railway (United States). Postgres on Supabase, region us-west-2 (United States). Cloudflare in front of the site. EU residency: not available yet.
Encryption
TLS in transit. Credentials and API keys sealed at rest with per-user envelope encryption (Fernet / AES-128).
Authentication
Password sign-in with rate limiting and breached-password screening against Have I Been Pwned. MFA and SSO: not yet available, post-launch.
Backups
Database backups are managed by Supabase. Our restore procedure has not yet been independently tested.
Penetration testing
No third-party penetration test yet. An internal red-team suite runs in CI on every change.
Vulnerability disclosure
Report to [email protected]. Machine-readable contact at /.well-known/security.txt.
Incident notification
CommitmentWe will notify affected customers by email within 72 hours of confirming a breach.
Data processing agreement
Draft published for review at /dpa; not yet executable pending legal review. To request the signed version once final, write to [email protected].
Data deletion & export
You can export your data, and you can request complete deletion, in line with your rights under GDPR and PIPEDA.
Certifications
None held. SOC 2 Type II is on our roadmap; the observation window starts after launch.
Company
LiquidCortex Inc., incorporated January 19, 2026 in Ontario, Canada. Toronto, Ontario. Canadian Business Number 744671033 RC0001.

Subprocessors

Third parties that process customer data on our behalf, and what each one handles.

RailwayHosts the application backend and Redis (United States).
SupabaseHosts the Postgres database (AWS us-west-2, United States).
CloudflareDNS, TLS termination and edge proxy in front of the site.
StripeBilling and payment processing. Card data never touches our servers.
NangoBrokers the OAuth connections you authorize to your own tools.
AnthropicModel provider. Receives task content for inference only.
OpenAIModel provider. Receives task content for inference only.
Together.aiModel provider. Receives task content for inference only.

Last updated: September 15, 2026

Talk to our security team

Evaluating LiquidCortex?

Detailed security documentation — architecture, controls, and our security questionnaire — is available under NDA for teams in evaluation. Email [email protected].

Found a vulnerability?

Report it to [email protected]. We acknowledge reports in good faith and ask for reasonable time to remediate before public disclosure.