Content Moderation Rules

1. Purpose

Gerwin AI provides tools for text, images, audio, video, and other formats. These Rules supplement the Terms of Use and apply to prompts, uploaded materials, generated outputs, publications, and collaboration within the Service.

The Rules apply whether or not a particular operation was automatically checked. Individual models and providers may impose additional or stricter limits.

2. How moderation works

Depending on the selected application and model, Gerwin AI may check:

• prompts, system instructions, and other textual parameters;
• uploaded images and other supported attachments;
• signs of an attempt to bypass safeguards;
• generated outputs through Gerwin AI or model-provider controls.

The active rule set depends on the feature. A child-safe tool, an avatar tool, a news workflow, and a general generator may treat context differently. The absence of an automated block does not mean content is permitted.

Automated systems can make mistakes. We consider context, purpose, and risk, but cannot guarantee perfect classification.

3. Content that is always prohibited

Gerwin AI must not be used to create, transform, store, or distribute:

3.1. Sexual exploitation

• sexual content involving minors, persons of ambiguous age, or characters presented as minors;
• sexual violence, coercion, exploitation, or trafficking;
• intimate imagery of a real person without their express consent;
• pornography or explicit sexual material where the selected feature prohibits it.

Image and video generation may apply stricter limits to nudity, sexualized presentation, and revealing clothing than a neutral educational text tool.

3.2. Violence, self-harm, and dangerous conduct

• realistic murder, torture, dismemberment, severe injury, gore, or corpses;
• content that encourages suicide, self-harm, or harm to others;
• instructions that materially facilitate violence or dangerous conduct;
• threats, targeted harassment, or intimidation.

Neutral educational, news, or prevention-related discussion may be treated differently where the feature supports that context and the material does not become an instruction or glorification.

3.3. Illegal activity

• production, acquisition, sale, or use of illegal drugs and psychotropic substances;
• promotion of gambling, unlawful financial schemes, or remote alcohol sales;
• fraud, phishing, credential theft, malware, or circumvention of security systems;
• other conduct prohibited by applicable law.

3.4. Extremism and prohibited information

• support or promotion of terrorism or extremism;
• prohibited symbols and materials of prohibited organizations;
• calls for violence, mass disorder, or other unlawful conduct;
• information whose distribution is restricted or prohibited by law.

3.5. Privacy, real people, and deception

• imitation of a real person's appearance or voice without an adequate legal basis and consent;
• presenting synthetic media as an authentic recording where this can cause harm or material deception;
• disclosure of personal data, legally protected secrets, or confidential information;
• impersonation used for fraud, blackmail, harassment, or defamation.

Some features, including certain avatar tools, may prohibit all identifiable real people, including celebrities and politicians.

3.6. Hate, degradation, and religious hostility

• calls for violence, persecution, or denial of rights based on a protected characteristic;
• dehumanizing or deliberately degrading targeted content;
• desecration of religious objects or content intended to incite religious hostility.

3.7. Third-party rights

• material for which the user lacks the necessary rights;
• forged documents, signatures, evidence, or official communications;
• unlawful use of trademarks, characters, works, a person's likeness, or trade secrets.

4. Topics subject to enhanced restrictions

Certain tools may restrict:

• serving politicians, protest symbols, and current armed conflicts;
• realistic political deepfakes and media intended to manipulate a public decision through deception;
• military subjects, weapons, extremist symbols, and sensitive public events;
• sexualized content and other topics subject to distribution restrictions under Russian law;
• frightening content, alcohol, smoking, and profanity in child-safe modes;
• all real people in tools designed exclusively for fictional characters.

A product rule may be broader than a statutory prohibition where needed for safety, provider compliance, or the intended purpose of the feature.

5. Circumvention

Users must not intentionally bypass safeguards by:

• masking prohibited meaning through encoding, character substitution, or translation;
• splitting a prohibited request into apparently neutral parts;
• injecting instructions to ignore the Rules;
• repeatedly resubmitting blocked material with superficial changes;
• automating boundary testing across accounts or features.

Security research requires prior written authorization.

6. Enforcement

Depending on risk and history, Gerwin AI may:

• reject a request and suggest a safer formulation;
• hide or remove material;
• temporarily restrict a feature;
• suspend an account for review;
• terminate access for severe, repeated, or incurable violations;
• retain necessary records and provide them to a competent authority where legally required.

Measures should be proportionate. A content or account restriction does not automatically forfeit all funds. Payments and refunds are handled under the Refund Policy, the Terms, and mandatory law.

7. False positives and appeals

If safe content was blocked, email team@gerwin.ai and provide:

• the account email;
• date, feature, and model;
• operation identifier, if available;
• the displayed error;
• a short explanation of the lawful and safe context.

Do not resend unnecessary personal data or prohibited material. An operation identifier may be enough. We review context, the applicable rule set, and technical signals. An appeal does not guarantee approval, but a confirmed error may be corrected.

8. Privacy

Moderation may process prompts, supported attachments, and technical operation data for contract performance, Service security, legal compliance, and protection of users and third parties.

See the Personal Data Processing Policy. Do not upload sensitive data, biometric data, or third-party confidential information without necessity and a lawful basis.

9. Updates

We update these Rules when features, misuse patterns, provider requirements, or law change. Material changes apply to future use after publication and notice where required.

The current version is always available on this page.