可接受使用政策 · Acceptable Use Policy

This Acceptable Use Policy ("AUP") forms an integral part of qihuaai.com (operated by Stellar Byte Tech Team / StellarByte, incorporated in British Columbia, Canada) and has the same force as the Terms of Service, Privacy Policy and Refund Policy.

The Platform provides AI generation tools, but you bear sole responsibility for all uploaded materials, input prompts and generated outputs. You must independently evaluate whether the generated content complies with the laws of your jurisdiction, the compliance requirements of your distribution channels, and the lawful rights of third-party rights holders (trademark, copyright, portrait, privacy).

We reserve all rights to take action against violating accounts, including but not limited to real-time blocking, content removal, rate limiting, warnings, suspension, permanent ban and disclosure of user information to law enforcement.

• Improper content involving minors: generation, distribution or retrieval of any sexual, sexualized or exploitative content involving minors (CSAM), whether real or AI-synthesized, regardless of any "artistic" or "educational" framing. On discovery: immediate permanent ban plus reporting to the Canadian NCECC (National Child Exploitation Crime Centre) and the U.S. NCMEC.
• Deepfakes of real persons without consent: generating or distributing deepfake portrait, voice or behavior content of real natural persons without written authorization, including but not limited to politicians, public figures and private individuals. Deepfake pornography results in permanent ban plus law-enforcement cooperation.
• Violence / terrorism content: planning, instructing or assisting the manufacture of conventional / chemical / biological / nuclear weapons, explosives or narcotics, including specific formulations and modification schemes; promoting terrorism, extremism or armed violence.
• Hate speech / racism / systemic discrimination: hate speech, harassment, threats of violence or doxxing based on race, ethnicity, nationality, gender, sexual orientation, religion, disability and similar protected characteristics.
• Politically sensitive content (PRC + international): generation of national leader likenesses, political-event reenactments, content advocating subversion of state power, election interference, fake political news. Content must comply with the PRC Interim Measures for the Management of Generative AI Services, the Provisions on the Administration of Deep Synthesis Internet Services, and the political-compliance requirements of any distribution jurisdiction.
• Disinformation / election interference: mass generation of fake news, fabricated evidence or public-opinion manipulation aimed at interfering with elections or major public affairs.
• IP infringement (known trademarks / copyrighted characters): prompts that explicitly request commercial use of known trademarks, Disney / Marvel / Studio Ghibli / Pokémon characters or similar copyright-protected likenesses.
• Phishing / fraud / malware: constructing phishing sites, forging credentials, impersonating others; distributing ransomware, spyware, keyloggers; drafting fraud or scam copy.
• Unauthorized intrusion / sanctions evasion: unauthorized access to others' systems; providing compute or inference services to entities or individuals sanctioned by the U.N., OFAC or Canada's SEMA; assisting money laundering.
• Platform abuse: evading billing, bypassing API key limits, mass top-up bonus harvesting, multi-account arbitrage, constructing adversarial examples to attack the moderation system, automated bulk spam requests.

• Your own portrait / voice: uploading your own photos or voice for generation (e.g. face swap, AI voice). You should declare "self-authorized" in your prompt or upload metadata.
• Authorized third-party likenesses / content: e.g. brand-authorized spokesperson likenesses, contracted models, IP partnerships. You should retain authorization records (contracts, emails, model agreements) for review on request.
• Adult content (NSFW): most upstream models disable NSFW; the Platform disables it by default. If an adult model is launched in the future under compliance review, it will be opened only through a separate KYC channel. At present this is treated as prohibited.
• Portrait / voice imitation practice: non-commercial imitation for learning purposes (e.g. portrait-painting practice generating celebrity likenesses) is not actively prohibited, but commercial distribution (social media uploads, advertising) still requires authorization.

• Celebrity-style imitation ("Van Gogh style", "Studio Ghibli style"): artistic style is not an object of IP protection, but the moderation team watches for intentional source confusion.
• Historical figures (Shakespeare, Napoleon, etc.): public-domain individuals, but insulting, obscene or politicized treatments may be manually rejected.
• Deceased celebrities: handled per jurisdiction (some U.S. states recognize descendible publicity rights; PRC Civil Code recognizes posthumous personality interests). We generally reject commercial advertising generations.
• News event reenactment: not categorically prohibited, but fabricated scenes or interviews are treated as prohibited disinformation.
• Violence / gore (for film and game production): acceptable in artistic and design contexts but not for real-world threats or incitement to violence.

The Platform applies multi-layer moderation to every invocation: (a) client-side keyword pre-check (real-time blocking of high-risk terms); (b) OpenAI moderation API (server-side prompt and upload review, synchronous block); (c) upstream provider internal moderation (asynchronous; upstream may reject with 422 MODEL_REJECTED under its own policies); (d) sampled manual review (triggered by alerts or reports).

Moderation outcomes: PASS / WARN (gray area, allowed but logged) / SOFT_BLOCK (suggest the user revise the prompt, no credits charged) / HARD_BLOCK (rejected outright and recorded as a risk score).

If you disagree with a moderation decision, you may appeal via abuse@qihuaai.com with the call_id and context. We commit to a first response within 24 hours and a decision within 3 business days.

• Tier 1 (minor): gray-area prompts triggering moderation, outputs containing minor violations. Action: warning + removal of the output + rate_limit halved for 24 hours + risk-score record.
• Tier 2 (moderate): e.g. three Tier-1 violations within 24 hours, or a single clear attempt to bypass moderation. Action: 24-72 hour account freeze, all invocations blocked, no refund of credits consumed during the freeze.
• Tier 3 (severe): any single violation of Section 2 ("Strictly Prohibited"). Action: immediate permanent ban + refund of unused credits (no retroactivity; already-consumed credits are not refunded) + entry on upstream provider blacklist + referral to law enforcement for criminal matters.
• Appeals: Tier 1 and Tier 2 actions can be appealed within 7 days via abuse@qihuaai.com. Tier 3 actions (especially involving CSAM, terrorism, or unauthorized deepfakes) are not subject to ordinary appeal; compliance review is only conducted where the user proves the account was compromised.

If you discover suspected violating content, accounts or invocations (qihuaai.com users or qihua-AI-watermarked outputs on third-party distribution channels), please email abuse@qihuaai.com immediately with: (a) link or screenshot of the violating content; (b) violation type (refer to the Section 2 list); (c) your relationship to the content (victim / informant / general reporter); (d) your contact information (anonymous reports are also accepted).

Response commitments: CSAM reports receive a first response within 4 hours plus removal and reporting to NCECC / NCMEC within 24 hours. Other categories receive a first response within 24 hours plus a decision within 3 business days.

Where disclosure is legally required (valid subpoena, court order, emergency safety request), we will cooperate in accordance with the laws of the applicable jurisdiction. Unless prohibited by law (e.g. criminal gag orders), we will notify the affected user where reasonable.

Abuse reports and security: abuse@qihuaai.com (24-hour SLA, 7x24 queue)

Compliance and law-enforcement requests: legal@qihuaai.com

Technical support: support@qihuaai.com

Moderation-decision appeals: abuse@qihuaai.com (subject prefix "Appeal" + call_id)