Built for authorised security work

The service is built for authorised security research. We add no provider-side filtering on top of the models, so these rules are yours to follow.

What counts as authorised

You may point the models at systems in one of these four situations.

Systems you own

Your own infrastructure, applications and devices.

A written engagement

A client contract or scope document that covers the systems and the methods.

A lab

An environment you built or were given for testing, isolated from real targets.

A contest

A CTF or bounty programme whose rules allow the work.

Rules for every instance

01

Authorised use

Use the service only for security research you are allowed to perform: systems you own, a written engagement, a lab, or a contest whose rules allow it.

02

Dual-use

The models can assist work that is legitimate inside an authorised test and harmful outside one. You stay responsible for which side you are on.

03

No attacks

Do not use an instance to gain access to, disrupt, or take data from systems, accounts or networks you are not allowed to test.

04

No malware or phishing

Do not use an instance to create or improve malware, exploits or phishing for anyone outside an authorised engagement.

Use remains bound to the terms and to applicable law.

In practice

Examples of work that fits, and work that does not.

Fits the rules

  • Penetration tests within a written scope
  • Vulnerability research on systems you may test
  • Exploit analysis and reverse engineering in a lab
  • Malware analysis in an isolated environment
  • Incident response, SOC analysis and code review
  • CTFs and contests whose rules allow it

Does not fit

  • Accessing systems, accounts or networks you may not test
  • Disrupting services or taking data without permission
  • Creating or improving malware for use outside an engagement
  • Building exploits or phishing for anyone outside an engagement

Verification for Red Team models

The more capable the Red Team model, the more we want to know who uses it. A model whose licence does not allow a Red Team variant is offered as Blue Team only.

EntryBlue Team: No verification requiredRed Team: Verified email and intended use
BasicBlue Team: No verification requiredRed Team: Verified email and intended use
ProBlue Team: No verification requiredRed Team: Full verification required
FrontierBlue Team: No verification requiredRed Team: Full verification, business customers only

The binding rules

This page explains the rules. The complete, binding text is the acceptable use policy, which is part of the terms of service.

Questions

Which systems may I test?

Systems you own, systems covered by a written engagement, a lab, or a contest whose rules allow it.

Do you filter what the models answer?

We add no provider-side filtering on top of the models. That is why the rules on this page apply to you, and why you stay responsible for how you use an instance.

What if I am not sure whether something is authorised?

Treat it as not authorised until the scope is confirmed in writing by the owner of the system.

Do all models need verification?

No. Blue Team variants need none. Red Team variants of Entry and Basic need a verified email, a payment method and a short description of your intended use. Pro and Frontier require full verification with an identity check, company details and a reviewed description of your intended use. Frontier is available to business customers only.

Where are the binding terms?

In the acceptable use policy, which is part of the terms of service. Use also remains bound to applicable law.

Request early access

Tell us about your team and the models you want to run. We will contact you as soon as your tier is available.