Acceptable Use Policy

Scope and how this policy works

This policy applies to the Website, to everything else Geocentric operates (our repositories and community spaces on GitHub, Hugging Face and similar platforms), and it states what we expect of anyone who uses a Geocentric model. It forms part of the Terms of Service.

It does not change a model's licence. Our open-weights releases are licensed under the Apache License 2.0, which grants rights without field-of-use restrictions, and this policy does not add any. What this policy does is tell you plainly what we consider unacceptable, commit us to what we will and will not do about it, and set the rules for our own platforms. You remain solely responsible for how you use a model, and unlawful use is unlawful whatever a licence says.

Use that is welcome

We do not want a policy that makes legitimate technical work feel risky. The following are expressly welcome:

• Programming, debugging, code review and refactoring.
• Reverse engineering and analysis of software you own or are otherwise authorised to analyse.
• Cybersecurity education, defensive security work, threat modelling and detection engineering.
• Vulnerability analysis and exploit research against systems you own or are authorised to test, in a way that stays within that authorisation.
• Malware analysis in controlled environments.
• Academic and independent research, including research about the model itself.
• Benchmarking, evaluation, red-teaming and publishing what you find, including results that make Geocentric look bad.
• Criticism of Geocentric, and direct comparisons with competing models.
• Automation and scripting for your own legitimate work, and accessibility work.
• Ordinary curiosity. Probing what a model will and will not say, or finding the edges of its behaviour, is normal research and not by itself a violation of this policy.

Use that is not acceptable

Do not use Geocentric models or services to do, plan or materially assist any of the following.

Against people

• Sexual exploitation or abuse of minors, or any sexualised depiction of a minor.
• Producing or distributing non-consensual intimate imagery.
• Stalking, harassment campaigns, doxxing or unlawful surveillance of individuals.
• Human trafficking or exploitation.
• Unlawful discrimination against people based on protected characteristics.
• Impersonating a real person or organisation in order to deceive or defraud, or generating deceptive content meant to manipulate elections or to pass as a person when it is not.
• Fraud, phishing, extortion or other deception for financial or personal gain.
• Encouraging or instructing self-harm or violence against others.

Against systems

• Unauthorised access to systems, networks, accounts or data.
• Theft or trafficking of credentials.
• Writing, deploying or distributing malware, ransomware or other destructive code to use against third parties, or building tools whose evident purpose is unauthorised access.
• Denial-of-service or other attacks intended to disrupt a service.
• Exploiting a vulnerability against a system you do not own and are not authorised to test, including turning a vulnerability a model helped you find into an attack.

Mass-casualty weapons

• Developing or acquiring biological, chemical, radiological or nuclear weapons, or attacks on critical infrastructure intended to cause serious harm.

Against our platforms

• Attacking, probing without permission, or degrading geocentricai.com or our repositories, other than good-faith research under our security policy.
• Spam, impersonation of Geocentric, or misuse of our names and logos as set out in Licences and trademarks.
• Automated scraping of the Website beyond ordinary personal use.

Security work: authorisation matters

Security research often looks the same as an attack from the outside. The line we draw is authorisation: you own the system, you have written permission from its owner, or it is a deliberately vulnerable target built for practice. If you do not have that, do not test it, and do not use a model to help you test it. Authorisation also has a scope: staying within it is part of staying within this policy.

We do not know and cannot verify who is using a downloaded model or for what. That is why we describe our expectations plainly, and why responsibility for authorised use rests with the person running the model.

Consequential decisions

Do not present Geocentric output as a professional determination, and do not use it as the sole or automatic basis for a decision that materially affects a person's employment, housing, lending or credit, insurance, education, health care, or access to government services or essential services.

People can and do ask AI systems about all of these subjects, and our models will answer. The line is between information and a decision made about someone. If you build a system that makes or substantially assists consequential decisions, that is a regulated activity in a growing number of jurisdictions, and it is your obligation to meet the applicable requirements. This policy does not do it for you. See our AI and Model Disclosure.

Safeguards

Our models carry only the safeguards described on their model cards, and some carry none. If you build on a model, you decide what safeguards your product needs and you are responsible for adding them. Do not remove or work around safeguards that a downstream provider has put in place to protect other people.

How we enforce this

We would rather correct a problem than remove someone. Where we act, we consider the context of the use, its severity, whether it looks intentional, whether it has recurred, and the risk it creates for other people.

Our practical controls apply to what we run: we can remove content from our repositories and community spaces, block or restrict access to the Website, and report serious cases to the platforms we publish on. We cannot recall a model that has already been downloaded or monitor how it is used, and we do not claim to. Serious cases, particularly child sexual abuse material, credible threats of violence or active attacks on infrastructure, may be reported to the appropriate authorities.

If you believe a restriction was applied in error, contact contact@geocentricai.com and we will look at it.

Reporting abuse

To report abuse of Geocentric models or services, contact contact@geocentricai.com with enough detail for us to investigate. Security vulnerabilities should go to our security contact instead. If someone is in immediate danger, contact local emergency services first.