Content Moderation Policy

This Content Moderation Policy explains how 2.71 keeps the Service safe: what we review, how we review it, and what happens when someone breaks the rules. It supplements the Terms of Service and should be read together with the Prohibited Content Policy and the Acceptable Use Policy.

1. Purpose and principles

2.71 exists for people to create synthetic images and video. That freedom depends on a small set of firm principles:

• Safety first. No creative use of the Service outweighs the safety of real people, especially minors.
• Zero tolerance for minors. Any attempt to sexualize a minor, real or fictional, ends in permanent removal from the Service. There is no warning tier for this.
• Proportionality. Outside of zero-tolerance cases, enforcement escalates with severity and intent, not with a single mistake.
• Consistency. The same conduct is treated the same way, whoever the account belongs to.
• Human oversight. Automated systems make the first pass, but consequential decisions, such as suspensions and bans, are checked by a trained person before they are final.
• Privacy of your Outputs. Your generations are private by default. We do not browse or review your private Outputs except to investigate a moderation flag, a report, or a legal or safety obligation. Public characters and content you choose to share are reviewed more broadly because other users can see them.

2. What we moderate

We review:

• Prompts submitted to generate images, video, or characters.
• Reference images you upload.
• Characters, including their names, descriptions, and any public listing.
• Outputs, particularly those attached to public characters or flagged by users or automated systems.
• Account behavior, such as repeated attempts to bypass safety systems.

We do not moderate based on your identity or your lawful preferences. We moderate content and conduct against the standards in the Prohibited Content Policy and the Acceptable Use Policy.

3. How we moderate

Moderation happens in layers:

• Automated pre-screening, before a generation runs. Prompts are checked by an automated content-policy gate, and reference images are checked for anyone who may appear under 18.
• Automated and human post-screening, after a generation completes. This includes automated checks on Outputs and human review of flagged content, public characters, and reported material.
• User reports. Anyone can flag content they believe violates our policies.
• Staff review. Trained reviewers examine flags, reports, and edge cases, and can remove content or act on accounts.
• Block records. Every blocked prompt or image is logged for enforcement and pattern detection, even though it is never shown to anyone.

4. Enforcement actions

Depending on severity, intent, and history, we may take one or more of the following actions:

• Rewrite a prompt automatically so borderline wording resolves to something compliant.
• Block a generation before it runs.
• Warning to the account.
• Content removal, including deleting an Output or character.
• Character de-listing, removing a public character from view while leaving it in the owner's private library.
• Temporary suspension of account access.
• Permanent ban, with any Coin balance forfeited and no refund.
• Referral to law enforcement and the National Center for Missing & Exploited Children (NCMEC), where content indicates apparent child sexual abuse material or credible sex-trafficking activity.

Ambiguous first-time issues typically result in a rewrite, a block, or a warning. Deliberate or repeated violations escalate to suspension or a permanent ban.

5. Immediate permanent ban

Some conduct results in an immediate permanent ban, with no warning tier and no refund of forfeited Coins:

• Any attempt to generate content that sexualizes a minor, including through misspellings, euphemisms, non-English wording, or anime and cartoon styles intended to depict a minor.
• Uploading a reference image of someone who appears to be under 18.
• Credible indicators of sex trafficking.

These cases are referred to NCMEC's CyberTipline, and where relevant to the National Human Trafficking Hotline (1-888-373-7888) or law enforcement, and we cooperate with any resulting investigation.

6. How you are notified

When a prompt or image is blocked, we show the reason in the Service at the time of generation. For account-level actions, such as a warning, suspension, or ban, we email the address on file, where we have one. We are not obligated to explain the specific signals behind an automated decision, since doing so can help bad actors evade the same systems in the future.

7. Appeals

If you believe an enforcement action was a mistake, you can appeal. Appeals are handled under our Complaint Policy: one appeal per decision, filed within 30 days of the decision, reviewed by someone who was not involved in the original call. We acknowledge appeals within 7 days. Content involving minors, sex trafficking, or other illegal activity is actioned within 24 hours of verification; other decisions are typically resolved within 30 days, and we will tell you if a case is taking longer.

8. Reporting content

If you see a public character, listing, or other content that you believe violates our policies, report it. See the Content Removal Policy for how to submit a report and what happens next.

9. Records and transparency

We keep a record of every block: the prompt or reference image, the reason, the account, and the time. These records are retained for as long as the account exists and afterward as needed to meet legal obligations, support investigations, and defend against fraud or abuse. We periodically review our moderation rules against new risks and update this policy as our models, our product, and the law change. Material changes are reflected in the date at the top of this page.

10. Our staff

Content flagged for human review is handled by trained staff. Access to moderation tooling and to flagged material is limited to the people who need it to do their job. Because this work involves difficult material, we take the wellbeing of our reviewers seriously, including limits on exposure and access to support.

11. Limits of moderation

No moderation system is perfect. Automated tools can block content that was actually fine, and in rare cases can miss content that was not. Where there is doubt, we generally err on the side of blocking rather than allowing content through, because the cost of a false negative is higher than the cost of a false positive. If you believe a block or removal was wrong, use the appeal process in Section 7.

12. Contact

Questions about this policy or how a specific decision was made can be sent to legal@271.dev.