Skip to content

Updated November 18, 2026

NASC security and privacy

NASC holds your team's AI work, so here is exactly how we look after it. This page is a plain summary. The Terms, Privacy Policy and Data Processing Agreement are the binding documents, and they win if anything here differs.

01

Your team owns its data

Your organization is the owner of everything it saves in NASC. We process it only to run the service for you. We do not sell it and we do not use it for advertising.

02

Nothing trains AI models unless you say so

Nothing your team saves is used to train AI models unless an owner turns Model improvement on. It is off for every new team. We do not allow the AI providers that power NASC to train on your content, and our AI gateway is set to zero data retention.

The AI tools you connect yourself, such as ChatGPT or Claude, follow their own vendors' terms.

03

What Scribe saves

Scribe is the small program that saves your AI sessions. It is designed to save only into a session a member has started by saying "nasc start", only from the AI tools an owner has allowed, and the member sees a confirmation that saving is on.

It saves the conversation, the AI's reasoning, and basic activity details: the time, the tool and platform, the computer name, the project and the paths of files worked on.

04

Keys and passwords in a chat

People paste secrets into AI chats by accident. Recognizable keys and credentials are removed from live sessions. No filter catches everything, a capture that cannot be read right away can be held unfiltered for up to 72 hours, and files you import are stored as you uploaded them. The safest habit is still to keep secrets out of AI chats.

05

Signing in

There are no NASC passwords to steal. You sign in with Google or with a one-time code sent to your email. Members can add an authenticator app. Sensitive actions, such as downloading an export or removing a member, make you prove again that it is you first.

06

Connected tools and computers

AI tools and Scribe connect with OAuth, the same standard used by "Sign in with Google". We never store a usable copy of a connection credential, only a one-way fingerprint of it. Credentials are short-lived and are replaced as they are renewed, and if a retired one is ever presented again the credential is revoked.

A computer that has been quiet for 14 days is signed out by a regular cleanup.

07

Encryption

Everything travels over encrypted connections (TLS), and the web app refuses unencrypted ones. Stored data is encrypted at rest by our hosting provider.

08

Separation between organizations

Every request is checked against your organization in the service itself, and the database enforces the same boundary again as a second layer. One organization cannot read another's saved work unless an owner has expressly granted a person from outside access.

A small number of named Archerion administrators have technical access to all organizations' data. They use it to run and secure the service, to help when you ask for support, and to answer legal requests.

09

Who on your team can see what

A workspace is shared. Every member can read the team's saved work, including conversations saved by teammates, so do not treat a saved session as private from your own team.

Roles (Owner, Admin, Dev Manager, Member, Viewer) decide who manages members, settings, capture and exports. Viewers can read and nothing else.

10

A record you can check

Saved sessions, decisions and transcripts are kept as the record. Summaries and wiki pages are views built on top of them, with page history and links back to the sessions they came from. When a newer decision replaces an older one, the older one stays in the history. If a summary is wrong, the evidence underneath it is still there.

The record changes only in set cases: recognizable secrets are removed, content is deleted when you ask, and a member who leaves is shown as "Former member".

11

What we log

Sign-ins, changes to admin settings, and data rights requests are recorded, along with who asked and why. Removing a member or exporting on someone's behalf requires a recorded reason.

12

Export and deletion

Owners and admins can export from the Control Panel. That export has size limits and does not include every record type.

Ask us in writing to delete specific content, a project or your whole workspace and it is done within 30 days, confirmed in writing, and gone from backups within 30 days after that. If you leave, you have 90 days to ask for a complete export before deletion. We keep our own account, billing and security records.

13

Backups

Our hosting provider takes managed database backups.

14

Where your data lives

Saved content is hosted in the United States (Northern Virginia) on Supabase, which runs on Amazon Web Services. Search indexes can be computed on short-lived servers in Germany that keep nothing once the job is done.

Web traffic passes through Cloudflare's global network, and AI providers process requests in a region the provider selects.

15

The companies we rely on

Supabase and Amazon Web Services (hosting and database), AWS Bedrock, Anthropic, Google Vertex AI and Vercel AI Gateway (AI processing), Hetzner (short-lived search indexing), Resend (email), Cloudflare (web delivery), Stripe (billing, which never sees saved content), Google (sign-in) and your browser's push service (notifications).

The full list with what each one does is in the Data Processing Agreement. Owners get 30 days' notice by email before we add or change one.

16

If something goes wrong

If we become aware of a breach involving your personal data, we tell you within 48 hours. To report a security problem, write to hello@archerionlabs.com and a person answers.

17

Health information

NASC is not built for protected health information under HIPAA, and our standard terms do not include a Business Associate Agreement. Do not save patient data in it.

18

What we do not have yet

We are a two-person company and we would rather tell you than have you guess.

  • No independent penetration test or security certification yet. Our reviews so far are our own.
  • No self-hosted version. NASC is a hosted service.
  • No uptime guarantee yet.
  • No customer-set retention period. Saved work is kept while your subscription is active.
  • No audit log that customers can browse, and we do not yet log each time an administrator reads data.

If one of these decides it for you, write to us. We will tell you where it sits on the list.

19

The binding documents

Questions people ask

Does NASC train AI models on our work?
No, unless an owner turns it on. It is off by default.
Where is our data stored?
In the United States, on Supabase and Amazon Web Services.
Is it encrypted?
Yes, in transit and at rest.
What does Scribe save?
Sessions a member starts with "nasc start", from tools an owner has allowed.
What if someone pastes an API key into a chat?
Recognizable secrets are removed before storage. Treat that as a safety net, not a guarantee.
Can my teammates see my saved conversations?
Yes. A workspace is shared with your team.
Can we delete everything?
Yes for everything your team saved: on written request, within 30 days, and from backups within 30 days after that. We keep our own account, billing and security records.
Can we take our data with us?
Yes. Owners and admins can export from the Control Panel, and if you leave you have 90 days to ask for a complete export.
Who are your vendors?
They are listed on this page and in the Data Processing Agreement.
Has NASC had an independent security test?
Not yet.
Can we self-host NASC?
No. NASC is a hosted service.
Will you tell us about a breach?
Yes, within 48 hours of becoming aware of it.
Can we use NASC for patient health information?
No.
NASC security and privacy · Archerion Labs