Moderation Policy

Effective Date: July 25, 2026
Last Updated: July 25, 2026

This policy explains why a prompt or Output might be blocked, how our safety systems work, and what to do when we get it wrong.

It is separate from our Acceptable Use Policy, which sets out what is prohibited. This document is about how those rules are enforced.

1. Why We Moderate

Generative media can be used to depict real people without their consent, to sexualize minors, to fabricate evidence, and to defraud. Moderation is how we make those outcomes less likely.

We aim to be restrictive where the potential harm is severe and irreversible, and permissive elsewhere. We would rather block some legitimate creative work than allow the categories of content that ruin lives.

2. How Moderation Works

Moderation happens at several points:

Before generation. Prompts and uploaded files are screened against our safety classifiers.
During generation. Third-party model providers apply their own filters.
After generation. Outputs may be screened before being returned or stored.
Across an account. We look at patterns of behaviour over time, not only individual requests.

Most moderation is automated, because the volume requires it. Decisions with serious consequences involve a human.

3. Why a Prompt Might Be Blocked

Common reasons include:

The prompt references a named real person, particularly a public figure
The uploaded image appears to depict a minor
The prompt requests sexual or intimate content
The prompt requests violence, gore, or self-harm content
The prompt appears designed to fabricate a real event or statement
The prompt or upload matches patterns associated with known abuse
The prompt appears crafted to evade a filter

Some blocks are contextual. The same prompt may be allowed in one context and blocked in another, for example where an uploaded reference image changes what would be produced.

We do not always explain the specific reason for a block. Detailed explanations make filters easier to defeat, which is precisely what the people we most want to stop are trying to do.

4. Provider Filters

Many models we offer are operated by third parties, and each applies its own safety filters.

A generation may be blocked by their systems rather than ours. When that happens we often receive only a generic rejection, and we may be unable to tell you more than that the provider declined the request. Trying a different model sometimes succeeds where a provider's filter was the cause.

5. Automated Systems Make Mistakes

Classifiers produce false positives and false negatives.

You may find a legitimate creative request blocked. Someone else may find that prohibited content slipped through. Neither means the system is absent; both mean it is imperfect.

Two consequences follow. First, if your work was wrongly blocked, appeal — we want to know. Second, a generation completing successfully is not a determination that the content is lawful or permitted. You remain responsible for what you publish.

6. Manual Review

We review content manually when a report is received, when automated signals suggest a serious violation, when an appeal is submitted, and when we are required to by law.

Reviewer access to your content is limited to what is needed for the review, and is logged. Reviewers are subject to confidentiality obligations.

7. Enforcement Actions

Depending on severity and history, we may:

| Action | Typically applied when | | --- | --- | | Block a single generation | A prompt trips a safety filter | | Warn the account | A first, minor, apparently inadvertent violation | | Remove content | Content violates our policies or is subject to a valid legal notice | | Restrict a feature | Repeated violations within one capability | | Rate-limit or throttle | Platform abuse or disproportionate load | | Suspend the account | Repeated or serious violations | | Terminate permanently | Severe violations, or persistent violation after warning | | Report to authorities | Where required by law, or where there is a risk to life or a child |

We aim to respond proportionately. Violations involving minors, non-consensual intimate imagery, or the non-consensual depiction of a real person result in immediate permanent termination without prior warning.

8. Credits and Blocked Generations

Credits are not consumed by a generation blocked before processing begins.

Where a generation is blocked by a provider after processing has started, we restore the credits consumed. If you were charged for a blocked generation and the credits were not returned, contact billing@inferon.ai with the generation identifier.

9. Appeals

If you believe a decision was wrong, contact legal@inferon.ai within 30 days.

Include:

Your account email
What was blocked, removed, or restricted, with the generation or asset identifier where you have it
Why you believe the decision was wrong
Where consent is at issue, evidence of the consent or rights you hold

We aim to acknowledge appeals within 2 business days and to decide within 10 business days. We will tell you the outcome and, where we can do so without undermining our safety systems, the reason.

Appeals are not available for terminations relating to content involving minors.

10. Human Review of Automated Decisions

Where a decision affecting your account was made by an automated system, you may request review by a person. Say so in your appeal.

We do not use fully automated decision-making to terminate accounts for anything other than confirmed matches against known illegal content.

11. Attempts to Evade Moderation

Deliberately working around safety systems is a violation in itself, whether or not it succeeds.

This includes prompt injection, obfuscated or coded phrasing, splitting a prohibited request across multiple generations, using one model to produce an intermediate result that another would have refused, and creating new accounts after a suspension.

Accounts engaged in systematic evasion are terminated, and evidence may be retained for enforcement and legal purposes.

12. Contact

Appeals and enforcement questions: legal@inferon.ai
Reporting content: abuse@inferon.ai, or see our Abuse Reporting Policy
Billing questions about blocked generations: billing@inferon.ai