Acceptable Use Policy

This policy sets out what MaddLabs and the Models may not be used for. It forms part of the Terms of Service and applies to the Customer, to anyone using its API Keys and to users of the Customer's products built on the API.

Scope and general principles

1. The Customer uses the API in compliance with the law, the Terms, the licences and usage policies of the Models and this policy.
2. A Customer that offers features based on the API to its own users puts in place rules in its product at least as strict as this policy and acts on violations.
3. The prohibitions apply both to sending requests and to using Model outputs.

Prohibited content and activities

The API must not be used to create, distribute or facilitate:

1. Illegal activities: in particular trafficking in drugs, weapons or people, money laundering, sanctions violations, infringement of intellectual property rights and other crimes.
2. Child sexual abuse material: without any exception, including fictional and generated content. MADD reports every case that comes to its attention to the competent authorities and terminates the agreement.
3. Non-consensual intimate imagery: sexual content depicting real people without their consent.
4. Hate and harassment: content inciting hatred or discrimination based on protected characteristics, harassment, intimidation and humiliation of others, and denial of documented genocides.
5. Violence and terrorism: threats, incitement to violence, glorification or support of terrorism, and instructions for causing serious physical harm, including making chemical, biological, radiological or nuclear weapons or explosives.
6. Self-harm: content encouraging self-harm, suicide or eating disorders, or instructing how to carry them out.
7. Fraud: phishing, scams, fake reviews, impersonation of people or institutions, and misrepresenting the identity or origin of content.
8. Deliberate disinformation: creating and spreading at scale content known to be false in order to mislead, including to influence elections or referendums.
9. Privacy violations: unlawful surveillance, collecting or disclosing personal data without a legal basis, using a person's likeness or voice without their consent, and identifying people from biometric data unlawfully.
10. Practices prohibited by Regulation (EU) 2024/1689 (the AI Act): in particular harmful subliminal or manipulative techniques, exploiting vulnerabilities of persons, social scoring and emotion recognition in the workplace or in education, subject to the exceptions in that regulation.
11. Advice posing as professional: presenting Model outputs as legal, medical or financial advice from a qualified professional without review by such a person, and making decisions of significant importance for people based on a Model output alone, without human oversight.
12. Security violations: creating malware, attacking computer systems, exploiting vulnerabilities without the system owner's permission and circumventing security measures.

Safety research and content moderation may require processing content from the categories above. This is permitted when the purpose is to protect against harm, not to cause it, and when done lawfully. If in doubt, write to domaai@madd.im before starting such work. This exception never applies to child sexual abuse material.

Abuse of the service

You must not:

1. attack or overload the API, including running load tests or penetration tests without MADD's consent;
2. circumvent the Limits, including by creating multiple Accounts, using other people's API Keys or spreading traffic to avoid limits;
3. share, sell or publish API Keys;
4. resell access to the Models in breach of their licences, or without a separate agreement with MADD where access is resold as a raw API;
5. attempt to access MADD's infrastructure, other Customers' data or MADD's providers' systems;
6. otherwise disrupt the services or mislead MADD about the Customer's identity or the purpose of use.

Obligations of Customers building products

1. Inform users that they are interacting with an artificial intelligence system and label generated content where the law requires it.
2. Ensure human oversight of decisions that may significantly affect people's rights, health, employment or financial situation.
3. Comply with the licences and usage policies of the Models, including attribution requirements and use restrictions set by their creators.
4. Have a legal basis for processing the personal data you send to the API.

Enforcement

1. MADD does not store or review request content. Violations may be detected from metadata, third-party reports, information from a Marketplace or providers, and requests from authorities.
2. Depending on the severity of a violation, MADD may: ask the Customer to stop, lower limits, block an API Key, suspend the Account or terminate the agreement under the Terms. MADD may report violations that may constitute a crime to the competent authorities.
3. If a Model's creator reports that use of the Model breaches its licence, MADD may restrict the Customer's access to that Model.
4. The Customer may provide explanations and file a complaint under the Terms.

Reporting violations

Violations, including content generated by a product built on the API, can be reported to domaai@madd.im. A report should describe the violation and, if possible, include the time, the model and details that help identify the Customer or product.

Changes

Changes to this policy are made in the same way as changes to the Terms.