Our approach
An organisation trusts Dienox with how it works, and trusts the agents on it to act within limits. So security is designed in, not added on: what an agent may not do does not exist for it, every change can be traced and undone, and one organisation’s knowledge never reaches another.
Organisations kept apart
- Every organisation has its own knowledge base, kept separately from every other.
- Every request is resolved to exactly one organisation, and every query is scoped to it.
- A connection from an AI tool belongs to one organisation. It cannot see or reach another, whoever asks.
Permissions and approvals
- Every action on every connection is read, write or irreversible, and switched on or off per agent.
- An action that is off is not in the agent’s toolbox, so it cannot be called, however the agent is asked.
- Irreversible actions wait for a person with the right role, and an approval counts once.
- People have roles per organisation: administrator, editor or reader.
Accounts and access
- Passwords are stored only as secure hashes, and login attempts are rate-limited.
- AI tools connect through OAuth: access tokens last hours, not months, and can be revoked at any time.
- Sensitive changes, such as managing members or issuing tokens, ask for your password again.
- Dienox staff reach the admin console only with a fresh password confirmation, and what they change there is logged.
Infrastructure
- Hosted in the Netherlands, with daily backups.
- All traffic is encrypted in transit, and the site sits behind Cloudflare’s protection against attacks.
- Strict security headers on every page, including a content security policy and no framing by other sites.
- Limits on every machine-facing interface, so a runaway AI tool meets a clear error instead of a flood.
- Dependencies are kept current, and the platform is reviewed for security as it grows.
A trail of everything
What was read and what was changed, by which person or agent, for whom and at what cost. Every change to a knowledge base is kept in its history and can be undone. Your knowledge is kept in open formats, so you can take all of it with you at any time.
In your own environment
The Enterprise edition runs the same platform on your own server, in your Azure or your AWS, with a local model where nothing may leave. Your knowledge, your data and your access then never leave your environment. Join the beta to talk about it.
Responsible disclosure
Found a vulnerability? We want to hear about it, and we will work with you to fix it. Mail security@dienox.com with what you found, how to reproduce it and what you think the impact is. Our machine-readable contact is at /.well-known/security.txt.
We ask you to
- report it as soon as possible, and give us reasonable time to fix it before telling anyone else;
- not access, change or delete more data than needed to show the problem, and never data of others;
- not disrupt the service: no denial of service, spam or social engineering of our staff or customers;
- not use physical attacks, or attacks on the systems of our providers.
We promise to
- reply within three working days, and keep you informed while we fix it;
- not take legal action against you for research that follows these rules in good faith;
- handle your report confidentially, and not share your details without your permission;
- name you as the finder once it is fixed, if you would like that.
Out of scope
- findings without a demonstrable security impact, such as missing best-practice headers on a static file;
- reports from automated scanners without a working proof;
- vulnerabilities in the AI tools you connect, which belong with their own providers;
- rate limits on the public beta form, and the absence of a lockout after failed logins.