Input must respect others' rights and protect secrets
Input must not contain state secrets or data affecting national security or public interests, or infringe intellectual property, portrait, reputation, honor, name, privacy, personal-information or other lawful rights. Before uploading commercial information, especially trade secrets, ensure you have processing rights and apply necessary desensitization.
Supporting quotations
4.2 When submitting Input, you must: (a) Ensure it does not contain state secrets or data affecting national security or public interests, and does not infringe any person's intellectual property rights, portrait rights, reputation rights, honor rights, name rights, privacy, personal information rights and interests, or other lawful rights and interests; (b) For personal information, obtain informed consent (or separate consent for sensitive information), or apply anonymization measures that comply with legal requirements; otherwise, please do not upload such information to the platform; Do not submit sensitive personal information of minors under 14. (c) For commercial information (especially trade secrets), ensure you have the right to process it and apply necessary desensitization before upload.
Content & IP Rights · Restriction or exception
The terms prohibit listed political, violent and other illegal content
You must not create or publish content opposing constitutional principles; threatening national security, unity, honor or interests; leaking state secrets or subverting state power; or disparaging heroes and martyrs or infringing their specified rights. The list also covers terrorism, extremism, ethnic hatred or discrimination, disruption of religious policies or promotion of cults, rumors disrupting social or economic order, obscenity, pornography, gambling, violence or crime incitement, insults, defamation, rights infringements and other content prohibited by applicable law. This describes the contract's list, not an independent determination of illegality.
Supporting quotations
4.4 You must not use the Service to create or publish illegal content: (a) Content opposing the basic principles of the Constitution. (b) Content endangering national security, leaking state secrets, subverting state power, or undermining national unity. (c) Content damaging national honor or interests. (d) Content that distorts, disparages, desecrates, or denies the deeds and spirit of heroes and martyrs, or infringes upon the names, portraits, reputation, or honor of heroes and martyrs by insult, defamation, or other means; (e) Content promoting terrorism or extremism. (f) Content inciting ethnic hatred, discrimination, or undermining ethnic unity. (g) Content undermining national religious policies or promoting cults. (h) Content spreading rumors or disrupting social or economic order. (i) Content that is obscene, pornographic, gambling-related, violent, murderous, terrorist, or crime-inciting content; (j) Content insulting, defaming, or infringing on others' rights. (k) Other content prohibited by applicable law.
Content & IP Rights · Restriction or exception
The harmful-content ban extends beyond the illegal-content list
Kimi's rules prohibit creating or publishing seriously mismatched exaggerated headlines; gossip, scandals or misdeeds; improper disaster or major-accident comments; sexual innuendo; blood, horror, cruelty or other physically or mentally disturbing content; group or regional discrimination; and vulgar, philistine or kitsch content. They also prohibit content inducing minors toward unsafe behavior, conduct violating social morality or bad habits, and other content adversely affecting the online ecosystem.
Supporting quotations
4.5 You must not create or publish harmful content: (a) Exaggerated titles seriously mismatched with content. (b) Gossip, scandals, or misdeeds. (c) Improper comments on natural disasters or major accidents. (d) Sexual innuendo or provocation. (e) Blood, horror, cruelty, or content causing physical or mental discomfort. (f) Inciting discrimination against groups or regions. (g) Vulgar, philistine, or kitsch content. (h) Content that may induce minors to imitate unsafe behaviors or behaviors violating social morality, or induce minors to develop bad habits; (i) Other content adversely affecting the online ecosystem.
Content & IP Rights · Company practice
AI identification covers generated or significantly modified content
Moonshot says it identifies AI-generated text, images, audio and video under the named identification measures and national standards for transparency and traceability. For these rules, Generated Content includes content significantly modified by AI. Explicit identifiers are perceptible text, sound or graphics such as watermarks or banners; implicit identifiers are metadata markers usable for tracing and regulatory oversight.
Supporting quotations
To improve transparency in AI processing, protect the public's right to know, and support content traceability, we identify Generated Content such as text, images, audio, and video generated through artificial intelligence technology in accordance with the Measures for the Identification of Artificial Intelligence-Generated Synthetic Content and relevant national standards.
(a) Generated Content: Content generated or significantly modified by AI, including text, images, audio, and video. (b) Explicit Identifier: Text, sound, or graphic prompts (such as watermarks or page prompt banners) directly perceivable by users. (c) Implicit Identifier: Technical markers embedded in file metadata that can be used for traceability and regulatory oversight.
Content & IP Rights · Restriction or exception
AI labels are added by default, with a limited explicit-label removal provision
Moonshot defaults to both explicit and implicit identifiers. An AI-Generated label appears prominently in the interface, generated images or other applicable places, and metadata identifiers are embedded under national standards without affecting normal use. Section 6.2 makes an exception for explicit identifiers when you use provided functions or have approval to remove them. Sections 6.3 and 4.7 otherwise prohibit removing, obscuring, altering, hiding or circumventing identifiers, potentially creating legal liability. The stated exception concerns explicit labels; it does not authorize removing metadata identifiers.
Supporting quotations
UNLESS YOU USE FUNCTIONS WE PROVIDE OR ARE APPROVED TO REMOVE EXPLICIT IDENTIFIERS, WE DEFAULT TO ADDING BOTH EXPLICIT AND IMPLICIT IDENTIFIERS TO ALL GENERATED CONTENT.
Explicit Identifiers: An "AI-Generated" identifier is displayed prominently in the interface, on generated images, or in other applicable positions.
Implicit Identifiers: WE EMBED IMPLICIT IDENTIFIERS IN ALL GENERATED CONTENT IN ACCORDANCE WITH NATIONAL STANDARDS. They do not affect normal use.
YOU MUST NOT REMOVE, OBSCURE, ALTER, HIDE, OR CIRCUMVENT ANY EXPLICIT OR IMPLICIT IDENTIFIERS ADDED BY THE SERVICE. DOING SO MAY RESULT IN LEGAL LIABILITY.
4.7 You must not circumvent security mechanisms: (a) Malicious circumvention, including but not limited to using variants, garbled text, characters, homophones, or other means to evade Service detection so as to input or generate illegal speech; (b) Malicious attacks, inducement, and poisoning by means such as identity impersonation, reverse induction, or jailbreak attacks; (c) Deleting, tampering with, or concealing the AI-generated content identifiers we have added (including prominent identifiers visible to the naked eye on Generated Content and hidden identifiers embedded in the metadata files of Generated Content through specific technical means), or engaging in other acts in violation of the Measures for the Identification of Artificial Intelligence-Generated Synthetic Content ; (d) Other conduct maliciously circumventing security management.
Cross-references
qualifies (model-proposed): Do not evade safeguards or tamper with AI-content identifiers — qualifies: The specific rule for provided functions or approved removal of explicit identifiers qualifies the general anti-tampering restriction; it does not authorize removing implicit identifiers.
Content & IP Rights · Restriction or exception
An AI label does not certify accuracy, ownership or suitability
Moonshot says an identifier is not its guarantee or endorsement of the content's authenticity, accuracy, ownership or applicability.
Supporting quotations
The identifier does not constitute our guarantee or endorsement of the content's authenticity, accuracy, ownership, or applicability.
Moonshot claims rights in the service subject to other lawful rights holders
To the extent permitted by law, Moonshot owns rights in the service, including copyright, trademarks, patents and other rights, except for rights belonging by law to other relevant rights holders.
Supporting quotations
5.1 Except for the relevant rights holders who are entitled to rights in accordance with the law, Moonshot AI owns all rights (including but not limited to intellectual property rights such as copyrights, trademark rights, and patent rights, and other rights) in the Service to the extent permitted by laws and regulations.
Content & IP Rights · User right or choice
You own input and output rights only within legal and contractual limits
The terms assign you intellectual-property and other rights in your Input and Output to the extent permitted by law, provided you comply with the agreement and do not infringe others' lawful rights. You remain responsible for Content. Similar output may be generated for others, and your output rights do not extend to their content. Your ownership coexists with the free model-optimization license granted to Moonshot, with a contact route if you do not want training use.
Supporting quotations
5.2 On the premise of not infringing others' lawful intellectual property rights or other rights and complying with this Agreement, and to the extent permitted by law, 5.2 you own the intellectual property rights and other rights in your Input and Output, and bear the obligation to be responsible for the Content. Due to technical limitations, we cannot guarantee that other users’ content is completely different from yours, and the occurrence of similar Output cannot be ruled out; accordingly, Your rights over the Output do not extend to other users' content.
5.3 TO IMPROVE THE SERVICE, YOU GRANT US A FREE LICENSE TO USE YOUR INPUT, OUTPUT, AND FEEDBACK FOR MODEL OPTIMIZATION WITHIN THE SCOPE PERMITTED BY LAW. IF YOU DO NOT WISH YOUR CONTENT TO BE USED FOR MODEL TRAINING, PLEASE CONTACT US THROUGH THE CONTACT INFORMATION PROVIDED IN THIS AGREEMENT.
Content & IP Rights · Restriction or exception
Handling illegal content is not promised as a universal monitoring duty
Moonshot says it may process illegal content under the terms, but that this does not itself create an obligation to do so. It cannot guarantee detection of all illegal acts.
Supporting quotations
9.6 We may process illegal content per this Agreement, but this does not constitute our obligation. We cannot guarantee detection of all illegal acts.
Content & IP Rights · Context
Content includes both your input and the model's output
The service uses large language models to generate Output, also called Generated Content, in response to your Input. References to Content cover both Input and Output.
Supporting quotations
We provide AI services powered by large language models. The model generates content in response ("Output" or "Generated Content") based on user input ("Input"). Input and Output are collectively referred to as "Content."