Usage Policy

MGS AI helps with code, debugging and everyday questions. This policy explains the few things it must not be used for, how requests are checked, and what happens when something is flagged. We stop clear abuse with general safety checks, without profiling you, and you can always appeal to a person.

1. What is not allowed

Only clear violations are flagged. When in doubt, nothing is flagged. Discussing, explaining, reporting on, preventing or writing fiction about a topic is fine unless it gives real, practical help to cause harm.

• Sexual content involving minors (100 strike points): Any sexual content involving minors, or sexualizing minors. The request is blocked and the account is suspended until a person has reviewed it.
• Biological, chemical, nuclear or radiological weapons (100 strike points): Real help to make, get or use biological, chemical, nuclear or radiological weapons. The request is blocked and the account is suspended until a person has reviewed it.
• Violent attacks and terrorism (5 strike points): Planning or helping violence against people, supporting terrorism, or making weapons to hurt people. The request is blocked.
• Malware and unauthorized intrusion (4 strike points): Building malware, ransomware, credential stealers or botnets, or attacking systems you do not own or are not allowed to test. MGS AI refuses.
• Non-consensual sexual content (4 strike points): Sexual content about real, identifiable people without their consent, including deepfake image prompts. Image prompts are blocked.
• Fraud and scams (3 strike points): Phishing kits or messages, scam scripts, fake documents, carding, or impersonation for fraud. MGS AI refuses.
• Harassment and threats (3 strike points): Credible threats, doxxing private people, or targeted harassment campaigns. MGS AI refuses.
• Hate and violent extremism (3 strike points): Promoting violence or dehumanization against protected groups, or extremist propaganda and recruitment. MGS AI refuses.
• Explicit sexual images (0 strike points): Explicit sexual image prompts about adults. Adult topics in text chat are not flagged. Only the image is blocked.

2. What is always allowed

• Security research, penetration testing of your own or authorized systems, CTFs, exploit explanations and defensive tooling.
• Malware analysis.
• Medical, drug-safety and harm-reduction questions.
• Crime fiction, history and news.
• Adult sexual topics in text.
• Jailbreak attempts and prompt-injection tests that ask for nothing harmful.
• Swearing.
• Questions about this policy.

3. How requests are checked

• The check is done by MGS AI itself, on our own servers, next to the chat model. Your text is never sent to another company for moderation.
• We check the latest message of each chat request (the text you typed, up to 8,000 characters; attachments are not scanned), every image prompt, and your custom instructions when you save them.
• If the check is slow or fails, the request goes through as not flagged.
• There are no hidden signals: no device fingerprinting, no tracking across sites and no linking of accounts by IP address.

4. What happens when something is flagged

Each flagged request adds the strike points of its category. Points count for 90 days. After each flagged request with points, the first matching rule applies:

1. Sexual content involving minors or weapons of mass destruction: the reply is stopped and the account is suspended until a person has reviewed it. The person lifts the suspension or makes it a permanent ban.
2. 20 points or more, or new points within 30 days after a 7-day suspension ended: the account is banned.
3. 12 points or more: the account is suspended for 7 days.
4. 6 points or more: the account is suspended for 24 hours.
5. Otherwise: you get a warning, and the account keeps working.

A suspension ends on its date by itself. Nothing steps down automatically in between, and the points stay until they expire. We email you about every warning, suspension and ban, with the category, the action and its end date, never with the text of your message. You can see your history in Settings.

5. Account states

• Warned: everything works. The app shows a notice with the category once; you can dismiss it.
• Suspended: you cannot send messages, generate images or upload files until the end date (or until the review). You can still sign in, read and export your chats, change settings, cancel billing and delete your account.
• Banned: the account is read-only. You can still sign in to export your data or delete the account. A running subscription is cancelled automatically, and new subscriptions are not possible.

6. If you are struggling

Messages about self-harm or suicide are never penalized and give no strike points. If MGS AI notices that you may be at risk, the app shows support resources:

• Latvia: Crisis and emotional support line, 116 123
• EU: Emergency services, 112
• UK and Ireland: Samaritans, 116 123
• United States and Canada: Suicide and Crisis Lifeline, 988
• Other countries: Find a Helpline, findahelpline.com

7. What we store

• Requests that are not flagged: nothing is stored about the check.
• A flagged request: your user id, the message id, where it came from (message, image prompt or custom instructions), the category, the strike points, the action, the version of the check, the time and the first 280 characters of the text so a person can review it. For self-harm no text is kept.
• The 280-character excerpt is deleted after 90 days. The rest of the record (without text) is deleted after a year.
• The engine never logs message text, including flagged text. Its logs hold only the request id, user id, category and severity.

More in the Privacy Policy.

8. Rate limits

The only signal that is not about content is the normal per-account rate limit. An account that hits rate limits more than 300 times in an hour is paused for 1 hour. This gives no strike points.

9. Appeals

Every suspension and ban can be appealed to a person: write to support@trymgs.ai from the email address of your account. We review the record and reply. Permanent bans for the most serious categories are always confirmed by a person.