Acceptable Use Policy

1. Scope

This Acceptable Use Policy (“AUP”) governs everything you do on the Pocket Expert AI platform at agentsdance.aiand in our desktop and mini-program clients (the “Service”), operated by Beijing AgentsDance AI Technology Co., Ltd. (北京跃迁效应人工智能科技有限公司), trading as Pocket Expert AI.

It applies to:

• everything you submit to the Service (prompts, files, URLs, connected data — “Input”);
• everything the Service produces for you, including generated images and video, text, slide decks, reports, code and web pages (“Output”);
• anything you publish, share or export from the Service, including published deliverables and agent presets you list in the agent store.

It applies whether the underlying model is billed through us or accessed with your own provider key as private model credentials. Using your own credentials does not exempt you from this AUP; your provider’s own policies apply in addition to it.

This AUP is incorporated into, and forms part of, our Terms of Service. Capitalised terms not defined here have the meaning given in the Terms. Violating this AUP is a breach of the Terms and may lead to the actions in Section 7.

2. Content Standards & Prohibited Categories

2.1 Absolutely prohibited — zero tolerance

You must never use the Service to generate, upload, store, request, publish or distribute:

• Child sexual abuse material (CSAM) and any sexualised depiction of minors — including real, drawn, animated or AI-generated imagery, and any content that sexualises a person who is, or appears to be, under 18. This also covers content that grooms, endangers or facilitates the abuse of minors.
• Non-consensual sexual content— sexual imagery of a real, identifiable person created or altered without that person’s explicit consent, including “nudify” / undressing tools, sexual deepfakes, and revenge pornography.
• Terrorism and violent extremism — content that promotes, celebrates, or provides operational support to terrorist or violent extremist organisations, and instructions for weapons capable of mass harm (explosives, chemical, biological, radiological or nuclear).
• Content that facilitates serious real-world violence — credible threats, incitement to violence against a person or group, or instructions intended to cause serious physical harm.

Content in this category is removed and the account is terminated immediately, without prior warning and without refund. Where required by law, we report it to the competent authorities and preserve the relevant records.

2.2 Prohibited content categories

You must not use the Service to generate, publish or distribute:

• Pornographic and NSFW content — sexually explicit images, video or text, including depictions of sexual acts, exposed genitalia, and content whose primary purpose is sexual arousal. Pocket Expert AI is not an adult-content platform; adult content is not permitted even between consenting adults and even where it would be lawful in your jurisdiction.
• Violent and graphic content — gore, mutilation, realistic depictions of death or torture, animal cruelty, and content that glorifies violence or self-harm, including material that encourages suicide, eating disorders or dangerous challenges.
• Hate speech — content that attacks, degrades, dehumanises, mocks, or incites hatred, discrimination or violence against a person or group on the basis of race, ethnicity, national origin, religion, caste, sex, gender, gender identity, sexual orientation, disability or serious disease. This includes hate symbols, slurs, and content that praises or denies atrocities such as genocide.
• Deepfakes and impersonation — synthetic images, video, audio or text depicting a real, identifiable person without their explicit consent; content that impersonates a person, brand, organisation or public authority; fabricated statements or events attributed to a real person; and synthetic media used to deceive, defraud, defame, harass, or manipulate elections. Consented creative or satirical use must be clearly labelled as synthetic.
• Intellectual property infringement — reproducing, imitating or distributing works, characters, trademarks, logos or trade dress that you do not own or have no right to use; prompts designed to reproduce a protected work; passing off Output as originating from a brand you are not authorised to represent; and counterfeit or pirated goods and media.
• Privacy violations— publishing another person’s personal data without a lawful basis, doxxing, biometric identification or surveillance of individuals without consent, and scraping personal data at scale.

2.3 General prohibitions

You must not use the Service to:

• break any applicable law or regulation, or breach card network rules;
• defraud, deceive or mislead — including phishing, fake reviews, fake identities or documents, pyramid and Ponzi schemes, unregistered financial solicitation, and academic dishonesty;
• build, distribute or operate malware, ransomware, credential harvesters, or tools for cyberattacks, or to probe, scan or attack any system without authorisation;
• send spam or bulk unsolicited messages, or automate abusive engagement on third-party platforms;
• present Output as the work of a qualified human professional in high-harm contexts — in particular medical, legal or financial advice — or otherwise conceal its AI origin where that would mislead someone to their detriment;
• circumvent security controls, rate limits, allowance limits or access controls, or attempt to defeat the safety measures described in Section 3;
• train or improve a competing AI or machine-learning model on Input or Output, or scrape Output at scale without our written permission;
• resell, sublicense or provide the Service to third parties except as expressly permitted by your plan.

3. Content Moderation

We are a small team and we are specific about what is automated and what is not, so that you know what to expect. Our moderation combines four layers:

3.1 Automated screening

• Platform-side prompt screening. Every image and video generation request is screened before it is sent to a model provider. Requests matching our zero-tolerance patterns (Section 2.1 — sexualised minors, and non-consensual sexualisation of real people) are refused, the user receives a policy notice, and the event is logged for review. This screening is deliberately narrow and high-precision so that legitimate creative work is not blocked.
• Model provider safety systems. Generation runs on third-party model providers, each of which applies its own safety classifiers to both the prompt and the generated output and refuses or blocks requests that violate their policies. Those refusals are surfaced to you as generation errors. This is the primary automated control for the broader categories in Section 2.2.
• Abuse signals. Every refused request is recorded against the account, so repeated attempts and attempts to evade screening are visible to us and bring the account into review.

We do not claim to run our own image or video classifier over every generated frame. Automated screening is a filter, not a guarantee — it will not catch everything, which is why the human layers below exist.

3.2 Human review

• Every automated block is recorded — with the account, the request and the reason — and is reviewed by a person, not closed automatically.
• Every report received under Section 5 is verified manually — we do not auto-close reports.
• Reviewers can access the specific Input and Output at issue for the purpose of the investigation, under the confidentiality and access controls described in our Privacy Policy.

3.3 Audits and spot checks

We carry out periodic spot checks on generation activity, and targeted audits of any account that has triggered a flag or a report. Content that is visible to other people — pages published from the Service and agent presets listed in the agent store — goes live without prior approval, and is prioritised in our spot checks and reviewed whenever it is reported.

3.4 Violation classification

4. Reporting Violations

Please report to us any content or behaviour on the Service that:

• breaches the content standards in Section 2 — in particular anything in Section 2.1;
• impersonates you, uses your likeness or voice without consent, or sexualises you without consent;
• infringes your copyright, trademark or other rights; or
• discloses your personal data without a lawful basis.

To let us act quickly, a report should include:

• a link to, or a copy of, the content at issue (a screenshot is fine), and where you encountered it;
• which category of this AUP you believe it breaches, and why;
• for rights complaints: what right you hold and evidence of it, plus a statement that you have a good faith belief the use is unauthorised;
• how we can reach you for follow-up.

Reports may be made anonymously, but we may be unable to give you an outcome if we cannot contact you. We do not disclose a reporter’s identity to the reported user unless required by law. Knowingly false or abusive reports may themselves lead to action under Section 7.

5. How to Report

• Email — [email protected]. This is a monitored mailbox and our primary reporting channel. Put REPORT in the subject line.
• Urgent matters — for anything involving child safety, terrorism, or an imminent risk of harm, email the same address with URGENT in the subject line. These are prioritised ahead of all other correspondence.
• Publicly hosted pages — to report a page published from the Service (a URL under agentsdance.ai/site/…) or an agent preset listed in the agent store, email the same address and include the URL or the preset name.

If you believe a crime has been committed, please also contact your local law enforcement. We cooperate with valid legal requests.

6. Response Times

We aim to acknowledge reports within 2 business days, and we target the following times to first substantive action, measured from when we become aware of the issue:

• L1 (critical): we aim to take first action within 24 hours and prioritise these ahead of everything else; critical reports are escalated the moment we see them, though action outside business hours may take longer;
• L2 (high): within 3 business days;
• L3 (medium): within 5 business days;
• L4 (low): within 10 business days.

These are the targets of a small team, set at a level we can actually meet rather than an aspirational number. Complex investigations may take longer; where they do, we tell you.

7. Enforcement Actions

Depending on the severity, frequency and intent of a violation, we may:

• remove or disable access to the content, and un-publish a deliverable or delist an agent preset;
• issue a warning and require correction;
• restrict specific features (for example image or video generation) on the account;
• suspend the account pending investigation;
• terminate the account and refuse future service;
• preserve records and report to law enforcement or other competent authorities where required by law.

We aim to act proportionately and to escalate for repeat behaviour, but for L1 violations we terminate immediately and without warning. Termination for a violation of this AUP does not entitle you to a refund of unused allowance; see our Refund Policy. Nothing in this section limits any other remedy available to us under the Terms or at law.

8. Appeals

If you believe an enforcement action was wrong, email [email protected] with APPEAL in the subject line, within 30 days of the action. Tell us what was actioned and why you believe it was mistaken, and include any context we would not otherwise have.

We respond to appeals within 10 business days. Where the original decision was automated, the appeal is reviewed by a person. If we got it wrong, we reverse the action and restore the content or account where we still can. Appeals against L1 terminations are reviewed, but content in that category is not restored.

9. Transparency & Changes to This Policy

We may update this AUP as the Service, the law, or the abuse landscape changes. The “last updated” date at the top of this page always reflects the current version. For changes that materially expand what is prohibited, we give at least 14 days’ advance notice, consistent with the notice provisions in our Terms of Service. Changes required to comply with law, or to address an urgent safety risk, may take effect immediately.

Continuing to use the Service after a change takes effect means you accept the updated policy.

10. Contact

This policy is operated by Beijing AgentsDance AI Technology Co., Ltd. (北京跃迁效应人工智能科技有限公司), trading as Pocket Expert AI. Reports, appeals and questions: [email protected].

Related documents: Terms of Service · Privacy Policy · Refund Policy.