Content safety
Acceptable Use Policy
Last Updated: July 13, 2026
LivRouter is committed to safe, lawful, and responsible AI access. This Acceptable Use Policy ("AUP") applies to all content generated, submitted, transmitted, stored, or processed through the Services and explains our content standards, moderation mechanisms, reporting process, response timelines, and enforcement actions. All users, developers, and organizations using the Services must comply with this AUP. It forms part of our Terms of Service.
1. Scope
- Text generated through chat, completion, code, document, or other language model features.
- Images, image edits, captions, OCR, or visual analysis when those features are available through the Services.
- Audio or video content, including speech, transcription, translation, voice, or media analysis when supported.
- User prompts, uploaded files, API requests, instructions, account activity, generated outputs, logs, and any downstream use of API keys.
2. Content Standards and Prohibited Categories
2.1 Zero Tolerance
- Child sexual abuse material (CSAM), sexual exploitation of minors, grooming, sextortion, or any sexual content involving a minor or a person who appears to be a minor.
- Terrorism, violent extremism, mass violence, credible threats, or instructions that enable serious physical harm.
- Instructions for weapons, explosives, biological, chemical, radiological, nuclear, or other mass-casualty harms.
- Incitement of genocide, ethnic violence, dehumanization, or severe hatred against protected groups.
- Non-consensual sexual deepfakes, intimate imagery, or synthetic media targeting a real person.
- Malware, phishing, credential theft, fraud, payment abuse, sanctions evasion, or attempts to compromise systems or accounts.
2.2 Restricted Content
- Sexual or NSFW content is not allowed unless expressly approved in writing and subject to appropriate age-gating, user consent, and applicable law. The current public service does not provide an approved adult-content workflow.
- Violent, bloody, or graphic content must have a clear educational, journalistic, safety, or creative context and may not glorify or encourage real-world harm.
- Political, medical, legal, financial, employment, housing, credit, or other high-impact content must not mislead users and must comply with applicable regulations.
- Synthetic media, impersonation, voice cloning, identity documents, or brand-like content must not deceive users about origin, consent, authority, or endorsement.
2.3 General Prohibited Conduct
- Generate, publish, or distribute false information, scams, spam, harassment, threats, or abusive content.
- Impersonate a real person, organization, platform, government body, brand, model provider, or rights holder in a misleading way.
- Infringe copyright, trademark, privacy, publicity, trade secret, or other rights, including counterfeit brand or product content.
- Represent AI-generated content as human-created where that would deceive users, customers, reviewers, or the public.
- Bulk-generate low-quality, deceptive, or manipulative content for search spam, social spam, fraud, or platform abuse.
- Use jailbreak prompts, prompt injection, obfuscation, automation, or other methods to bypass safety filters, rate limits, billing controls, or model restrictions.
3. Content Moderation Mechanisms
We use a layered moderation system combining automated review, human review, periodic checks, and provider-level safeguards. We may review both inputs and outputs where needed to enforce this policy.
3.1 Review Methods
- Automated review: prompts, requests, uploaded content, outputs, and usage patterns may be scanned using provider-level safety tools, keyword and semantic rules, anomaly detection, and risk scoring.
- Input filtering: prompts and instructions may be blocked, limited, or routed for additional review when they indicate prohibited intent or high-risk behavior.
- Output filtering: generated outputs may be blocked, withheld, or reviewed when they appear to violate this policy or provider safety requirements.
- Human review: reported content, flagged requests, payment risk, repeated suspicious activity, appeals, and severe safety categories may be reviewed by trained personnel under confidentiality obligations.
- Periodic checks: we may sample account activity, review enforcement quality, and adjust moderation rules to improve accuracy and reduce abuse.
3.2 Risk Classification
| Level | Description | Review | Action |
|---|---|---|---|
| L1 Severe | Child safety, terrorism, mass violence, severe abuse, malware, or illegal high-harm activity | Automatic block plus priority human confirmation | Immediate suspension, termination, preservation, and reporting where appropriate |
| L2 High risk | Serious policy violations, repeated attempts, fraud, impersonation, or high-risk deceptive content | Automatic flag plus priority review | Removal, feature restriction, account suspension, or payment review |
| L3 Medium risk | General prohibited content, harassment, misleading content, or rights concerns | Human review when flagged or reported | Warning, content removal, API key restriction, or model access limitation |
| L4 Low risk | Minor or first-time issues, unclear context, or incomplete reports | Report-triggered review or sampling | Warning, request for correction, or education |
4. Reporting Abuse or Policy Violations
- Content or activity that violates this policy, our Terms, applicable law, or third-party rights.
- Urgent child safety, terrorism, credible violence, account compromise, payment abuse, fraud, malware, phishing, or security issues.
- Impersonation, deceptive AI content, copyright or trademark infringement, privacy violations, or misuse of personal likeness.
- Attempts to bypass safety systems, resell access improperly, or hide the identity of actual end users.
4.1 Reporting Principles
- Confidentiality: we treat reporter identity and report details as confidential except where disclosure is needed to investigate, enforce this policy, protect users, or comply with law.
- Neutral review: reports are reviewed based on available evidence, policy requirements, and severity rather than the identity of the reporter or reported user.
- Abuse prevention: malicious, fraudulent, or knowingly false reports may be recorded and may lead to account action.
5. Reporting Channels
- Dedicated abuse and policy reports: [email protected] with the subject line "AUP Report".
- Child safety or urgent harm reports: [email protected] with the subject line "Urgent Safety Report" or "Child Safety Report".
- Copyright, trademark, or impersonation reports: [email protected] with the subject line "IP Report" or "Impersonation Report".
- Appeals: [email protected] with the subject line "AUP Appeal".
5.1 What to Include in a Report
- The account email, user name, API key name, request ID, URL, screenshot, timestamp, or other identifier related to the issue.
- The policy category involved, such as CSAM, NSFW, violence, hate, impersonation, copyright, trademark, fraud, or security abuse.
- A concise description of what happened and why it may violate this policy.
- Supporting logs, messages, links, files, or screenshots that help us verify the issue.
- A contact email for follow-up, unless you are reporting through an authenticated support channel.
6. Response and Resolution Timelines
| Violation Type | Initial Response | Target Resolution |
|---|---|---|
| L1 Severe | Within 2 hours | Within 24 hours |
| L2 High risk | Within 24 hours | Within 3 business days |
| L3 Medium risk | Within 3 business days | Within 7 business days |
| L4 Low risk | Within 5 business days | Within 15 business days |
Review flow: report received, severity classified as L1-L4, automated and human review performed as appropriate, enforcement decision made, action executed, and result communicated when a response channel is available. If more time is needed, we may notify the reporter where practical.
7. Enforcement Actions
| Enforcement | Measure | Typical Use |
|---|---|---|
| 1. Warning | Written notice requiring correction or deletion | First-time or minor L4 issues |
| 2. Content removal | Removal, refusal, or blocking of violating requests or outputs | Confirmed L3 or L4 violations |
| 3. Feature restriction | Reduced rate limits, disabled API keys, or restricted model access | Repeated minor violations or a single L3 violation |
| 4. Account suspension | Temporary suspension pending review | L2 violations or accumulated violations |
| 5. Permanent termination | Permanent loss of access to the Services | L1 violations, repeated L2 violations, fraud, or serious abuse |
| 6. Legal or provider escalation | Reports to service providers, payment processors, rights holders, or law enforcement where appropriate | Suspected illegal L1 activity or legally required disclosures |
8. Appeals
- If you disagree with an enforcement decision, you may appeal by emailing [email protected] with the subject line "AUP Appeal" within 30 days after receiving notice of the action.
- Include the account email, action taken, relevant request IDs or screenshots, and a clear explanation of why you believe the action should be changed.
- Appeals are reviewed by a person who was not the sole decision-maker for the original action when feasible. We aim to complete appeal reviews within 10 business days.
9. Transparency and Policy Updates
- We may use internal abuse metrics, provider notices, user reports, payment risk signals, and review outcomes to improve our safety systems.
- We may publish or summarize content-safety trends when appropriate, while protecting user privacy and security.
- Material changes to this policy will be posted on this page. Continued use of the Services after an update means you accept the updated policy.
10. Contact Us
- Content safety, abuse reports, urgent safety reports, intellectual property reports, and appeals can be sent to [email protected].
- General questions about this AUP can also be sent to [email protected].