AI-Generated Content & IP Notice

This notice explains the nature and limitations of AI-generated outputs from Acosmi services, how ownership of inputs and outputs is allocated, our content-labeling obligations, and how to report intellectual-property concerns.

Applicable region: International (users outside mainland China). Users in mainland China should switch to the Simplified Chinese (中国区) version.

Effective: June 3, 2026 | Last updated: June 3, 2026

1. Nature and Limitations of Generative AI Output

1.1 What generative AI output is

The Acosmi platform services operated by the Xiezhua group — including Acosmi (web agent and Agent Studio), Crab Code (desktop and CLI agent), Bench (众律宝 legal workspace), and Trusted Timestamp (可信时间章 e-signature and timestamping) (collectively, the "Platform" or "Platform Services") — use large language models (LLMs) and other generative AI technologies to produce text, code, images, video, audio and other content (collectively, "AI Output").

Generative AI models operate on probabilistic inference over statistical patterns learned from training data. They cannot guarantee that any output is accurate, complete, up-to-date, or free from error. You must understand and accept the following inherent limitations before relying on AI Output:

1.2 Not professional advice

AI Output does not constitute legal, medical, psychological, financial, investment, tax, accounting, or other regulated professional advice. Always consult a suitably qualified and licensed professional before making decisions in regulated domains.

Specific note regarding the Bench (众律宝) legal workspace: contract drafts, legal analyses, clause explanations, and other materials generated by Bench AI are provided for reference purposes only and are not a substitute for legal advice from a licensed attorney on your specific matter. Hongshen (Beijing) Legal Consulting Co., Ltd. provides lawful legal consulting services, but text automatically generated by AI does not itself constitute attorney work product. For matters with significant legal consequences, retain a licensed attorney for a formal legal opinion.

1.3 Your verification responsibility

When using AI Output, you are solely responsible for independently verifying its accuracy before relying on it, and you bear all risk and liability arising from your use of it. Platform Services are provided "as is." To the maximum extent permitted by applicable law, we make no express or implied warranties regarding the accuracy, completeness, fitness for a particular purpose, or non-infringement of any AI Output.

2. Ownership of Input Content

2.1 You retain rights to your inputs

Text, files, images, code, and other materials you submit to the Platform ("Input Content") remain the property of you or the applicable rights holder. Your submission of Input Content does not transfer ownership of that content to us.

2.2 Your representations

By submitting Input Content, you represent and warrant that:

• You own, or have all necessary rights, licences, and permissions in, the Input Content;
• The Input Content does not infringe or misappropriate any third party's copyright, patent, trademark, trade secret, right of publicity, defamation right, privacy right, or other proprietary or legal right;
• If the Input Content includes personal data of third parties, you have obtained all required consents and are in compliance with applicable privacy laws, including the EU General Data Protection Regulation (GDPR) where applicable;
• The Input Content does not contain Prohibited Content (see Section 6).

If your representations prove inaccurate and the Platform or any third party suffers loss as a result, you shall be liable accordingly.

2.3 Processing licence

To provide the Platform Services to you, you grant us a limited, non-exclusive, revocable, royalty-free licence to store, process, transmit, and use your Input Content to the extent necessary to deliver the requested service, including transmission to underlying model provider APIs. This licence is used solely to provide the service to you and does not include any right to use your Input Content to train foundation models (see Section 4). For further details on data handling, see our Privacy Policy.

3. Ownership of Output Content

3.1 General principle

To the extent permitted by applicable law, intellectual-property rights in AI Output generated at your direction through your use of the Platform ("Output Content") are assigned to and vest in you, the user who generated that output. We make no claim of intellectual-property ownership over Output Content.

3.2 Important caveats

You should be aware of the following significant limitations before using Output Content:

Identical or similar outputs for others. The same or similar prompts submitted by different users may produce substantially similar or identical Output Content. We cannot guarantee exclusivity, and we accept no liability for intellectual-property disputes arising from similar outputs being generated for multiple users.

Third-party rights may subsist. AI models are trained on existing works. Output Content may unintentionally be substantially similar to, or derived from, third-party copyrighted material, potentially giving rise to infringement risk. It is your responsibility to conduct any necessary clearance review before commercially exploiting Output Content.

Copyrightability of AI Output is uncertain and jurisdiction-specific. Copyright law in many jurisdictions, including the United States and EU member states, traditionally requires human authorship as a precondition for protection. Current legal developments include:

• Where you exercise substantial creative control over the final output — for example by making material creative choices in the prompt, or by editing and arranging AI Output — the resulting work may be protectable to the extent of your human creative contribution.
• Purely AI-generated content, without material human creative input, may not be eligible for copyright protection in certain jurisdictions.
• The position continues to evolve through litigation, guidance from copyright offices, and legislative proposals; consult qualified IP counsel for advice on your specific use case.

You bear responsibility for legal compliance of outputs. You are solely responsible for ensuring that your publication, distribution, and use of Output Content complies with all applicable laws, including the content-labeling obligations described in Section 5.

3.3 Special note: Bench legal document outputs

Ownership of AI-assisted legal documents (contract drafts, memoranda, pleadings, etc.) generated through Bench depends on the degree of creative and substantive reworking performed by the attorney or user, and on the law of the relevant jurisdiction. Before use in formal legal proceedings, such documents should be reviewed, verified, and adopted by a licensed attorney.

4. Model Training Commitment

4.1 We do not train on your inputs or outputs without separate consent

We commit that we will not use your Input Content or Output Content to train, fine-tune, or otherwise improve foundation AI models without obtaining your separate, specific, opt-in consent.

"Foundation models" means general-purpose pre-trained models operated by us or by third-party model providers.

4.2 Limited use for service improvement

We may use aggregated, de-identified statistical data derived from Platform usage to improve system performance, reliability, and safety mechanisms (such as abuse-detection classifiers). Such use does not constitute foundation-model training and is subject to our Privacy Policy.

4.3 Third-party model providers

The Platform integrates APIs from third-party model providers, including but not limited to Anthropic, Alibaba, Baidu, ByteDance, and Moonshot. These providers process data pursuant to their own privacy policies and terms of service. We contractually require our model provider partners to commit not to use Platform-user data for model training, but we cannot warrant compliance beyond such contractual obligations. Please refer to the Third-Party Data Processing section of our Privacy Policy for provider-specific details.

5. Content-Labeling Obligations

5.1 EU AI Act transparency requirements (international users)

The EU Artificial Intelligence Act (Regulation (EU) 2024/1689, applicable from August 2026 for relevant transparency obligations) requires providers of general-purpose AI and deployers of AI systems to take measures that ensure transparency to end users. Relevant obligations applicable to the Platform include:

For AI-generated content subject to Article 50 transparency obligations:

• Where the Platform provides a labeling feature for a given content type, it labels AI-generated outputs to inform users when they are interacting with or consuming AI-generated content, to the extent technically feasible.
• Where the Platform generates synthetic audio, image, video, or text content likely to be distributed to the public and a watermarking or metadata feature is available for that content type, we apply machine-readable disclosures (e.g., embedded metadata or watermarks) identifying the content as AI-generated.
• Users are informed when they are interacting with an AI system that is not a human.

User obligations. When you publish or distribute AI Output to third parties or the public — particularly content that could be mistaken for authentic human-created work — you are responsible for applying appropriate disclosures informing audiences that the content was generated or substantially assisted by AI. Removal of machine-readable markers or watermarks applied by the Platform may be unlawful under applicable law.

5.2 Deepfake and synthetic-media disclosure (EU AI Act Art. 50(4))

Under the EU AI Act and related member-state laws:

• AI systems that generate or manipulate images, audio, or video bearing a person's likeness (deepfakes) must label the output as artificially generated or manipulated.
• You must disclose that any deepfake content you create through the Platform and distribute publicly is AI-generated, unless you have the subject's explicit consent and the content serves a clearly artistic, satirical, or documentary purpose with appropriate context.

Failure to apply such disclosures may constitute a violation of the EU AI Act or national implementing legislation and may carry significant fines (up to EUR 15 million or 3% of worldwide annual turnover for certain infringements).

5.3 DMCA safe harbor context

With respect to the United States, the Platform operates as an online service provider and relies on the safe-harbor provisions of Title II of the Digital Millennium Copyright Act (17 U.S.C. § 512) (DMCA). Our notice-and-takedown procedures (Section 9) are designed to comply with DMCA safe-harbor requirements.

6. Prohibited Uses

6.1 Prohibited conduct

The following uses are expressly prohibited. Violations may result in immediate suspension or termination of your account and, where required by law, reporting to relevant authorities:

6.2 Legal consequences

Violations may implicate, among other laws:

• The EU AI Act (prohibited AI practices, transparency violations)
• The General Data Protection Regulation (GDPR) for privacy violations
• The Computer Fraud and Abuse Act (CFAA) and equivalent national cybercrime laws
• Copyright infringement statutes, including the DMCA and Directive 2001/29/EC
• Defamation, harassment, and other applicable tort laws
• Criminal laws relating to fraud, threats, child exploitation, and incitement

We will cooperate with law-enforcement investigations and may disclose information about your account where required by law or valid legal process.

7. Third-Party and Open-Source Model Notice

7.1 Model provider diversity

The Platform integrates APIs from multiple model providers to offer a diverse range of capabilities. Different models have different capability boundaries, training data, safety policies, content filters, and output characteristics.

7.2 Third-party model terms

When you use Platform Services, your requests are routed through to the applicable model provider's infrastructure. The usage policies and content policies of those model providers may impose additional restrictions on specific use cases and may be updated independently of the Platform's own terms. By using the Platform, you agree that such downstream restrictions apply to your use through the Platform.

Principal model providers (non-exhaustive; please refer to each provider's current policies):

7.3 Open-source model notice

Some features or private-deployment configurations may utilise open-source base models (such as openly licensed versions of Llama, Qwen-open, and similar). Those models are distributed under open-source licences (e.g., Meta Llama Community License, Apache 2.0) that may impose restrictions on commercial use, redistribution, or derivative works. If you use Platform API output commercially, you should review the applicable model licence terms for your use case.

8. Platform Intellectual Property

8.1 Platform ownership

The Platform's software code, user-interface design, icons, brand elements, training methodologies, proprietary model fine-tunes, API specifications, documentation, and website content (collectively, "Platform IP") are owned by or licensed to Xiezhua (Beijing) Intelligent Technology Co., Ltd. and/or Hongshen (Beijing) Legal Consulting Co., Ltd. (together, the "Xiezhua group"), and are protected by applicable intellectual-property laws worldwide.

8.2 Trademarks

The following trademarks and brand identifiers are owned by the Xiezhua group and may not be used without prior written authorisation:

• Acosmi and the Acosmi logo
• Crab Code (蟹爪) and associated graphic marks
• Bench (众律宝) brand and associated marks
• Trusted Timestamp (可信时间章) brand and associated marks

8.3 Restrictions on Platform IP

Without our express written authorisation, you may not:

• Copy, modify, distribute, or create derivative works based on the Platform's user interface or documentation;
• Reverse engineer, decompile, or disassemble the Platform's software;
• Use the Platform's trademarks, trade names, or service marks, or any confusingly similar identifiers, in a manner likely to cause public confusion;
• Remove, alter, or obscure any copyright, trademark, or other proprietary notices in or on the Platform.

9. Infringement Complaints and Notice-and-Takedown

9.1 How to submit an infringement notice

If you believe that content hosted on or generated through the Platform infringes your copyright, trademark, right of publicity, or other intellectual-property right, you may notify us under the DMCA (for US rights holders) or equivalent applicable law by submitting a written complaint to:

Email: fuwu@acosmi.com (please include "IP Infringement Notice" in the subject line)

Phone: 4000269678 (available Monday–Friday, 09:00–18:00 Beijing Time)

We will process complaints directed to the above designated agent in accordance with Section 512(c)(2) of the DMCA and equivalent international procedures.

9.2 Required contents of a valid notice

To be actionable, your infringement notice must include:

1. Rights-holder identification: Your full legal name (or the rights holder's name if you are acting as an authorised agent), mailing address, telephone number, and email address.
2. Identification of the protected work: A description sufficient to identify the copyrighted work, trademark, or other right you claim has been infringed (e.g., registration number, URL to the original work, or clear description).
3. Identification of infringing material: Information sufficient to locate the allegedly infringing material on the Platform, including the URL or other specific identifier of each piece of content.
4. Statement of good faith: A statement that you have a good-faith belief that the use of the material in the manner complained of is not authorised by the rights holder, its agent, or the law.
5. Statement of accuracy: A statement, made under penalty of perjury, that the information in the notification is accurate, and that you are the rights holder or are authorised to act on behalf of the rights holder.
6. Signature: Your physical or electronic signature (or that of a person authorised to act on your behalf).

For trademark notices, replace items 2–3 with: identification of the registered mark and the infringing use, and explain how the use is likely to cause consumer confusion.

We reserve the right to reject notices that materially fail to comply with these requirements.

9.3 Processing timeline

Misrepresentations in DMCA notices may expose the submitter to liability under 17 U.S.C. § 512(f). We reserve the right to take action against abusive or bad-faith notices.

9.4 Counter-notice procedure

If your content was removed following an infringement notice and you believe the removal was in error or the result of misidentification, you may submit a counter-notice. A valid counter-notice must include:

1. Identification of the material that was removed and the location at which it appeared before removal;
2. A statement, under penalty of perjury, that you have a good-faith belief that the material was removed or disabled as a result of mistake or misidentification;
3. Your name, address, and telephone number, and a statement that you consent to the jurisdiction of the federal district court for the judicial district in which your address is located (or, if outside the US, any judicial district in which we may be found), and that you will accept service of process from the person who provided the original notification;
4. Your physical or electronic signature.

Upon receipt of a valid counter-notice, we will forward it to the original complainant. If the complainant does not file a court action within 10 business days of our forwarding the counter-notice, we may restore the removed content at our discretion.

9.5 Repeat-infringer policy

In accordance with 17 U.S.C. § 512(i) and general best practices, we maintain and implement a policy to terminate, in appropriate circumstances, the accounts of users who are repeat infringers. "Repeat infringer" means a user against whom we have received multiple valid infringement notices that were not successfully contested via counter-notice. Consequences may include content removal, feature restrictions, account suspension, and permanent account termination.

10. Account Deletion

10.1 Deletion channel

You may request deletion of your account at any time. Account deletion is currently handled through customer support following identity verification; a self-service deletion path is being rolled out. Contact us at fuwu@acosmi.com or 4000269678, or refer to the Account Deletion Guide for the latest channel information. Once a deletion request is confirmed, the account enters a cooling-off period of approximately 7 calendar days to guard against accidental requests. You may cancel during this period. Once the cooling-off period expires, the deletion process becomes irreversible.

10.2 What happens to your data on deletion

When account deletion is completed:

• Your account profile, conversation history, AI Output history, uploaded files, and other stored personal content will be deleted or anonymised in accordance with the retention schedules set out in our Privacy Policy.
• Deletion does not extinguish any legal liability you incurred prior to deletion.
• Unused subscription time and membership benefits are handled in accordance with the refund provisions of our Payment, Subscription & Auto-Renewal Agreement.
• Your previously registered email address or phone number may be re-used to register a new account.

10.3 Content that cannot be deleted on request

Records of electronically signed contracts, Trusted Timestamp certifications, and legal filings created through Bench are subject to statutory record-retention obligations and may need to be retained for legally required periods regardless of account deletion.

AI Output and conversation history are not recoverable after deletion. Please download and save any content you wish to retain before initiating deletion.

10.4 Full details and assisted deletion

For the complete account-deletion process, see the Account Deletion Guide. If you encounter issues with self-service deletion or wish to request assisted deletion, contact us:

• Email: fuwu@acosmi.com (please include "Account Deletion" in the subject line)
• Phone: 4000269678 (Monday–Friday, 09:00–18:00 Beijing Time)

11. Changes to This Notice

11.1 Our right to update

We may update this notice at any time to reflect changes in technology, law, regulation, or business practice. The updated notice will be published on this page with a revised "Last updated" date.

11.2 Notification of material changes

For changes that materially affect your rights or obligations, we will provide at least 15 days' advance notice using one or more of the following methods:

• An in-Platform notification banner or modal;
• A notice sent to the email address associated with your account.

11.3 Continued use constitutes acceptance

Your continued use of the Platform after the effective date of any updated notice constitutes your acceptance of the changes. If you do not agree to the revised notice, you may delete your account as described in Section 10.

12. Contact Us

Operated by: Xiezhua (Beijing) Intelligent Technology Co., Ltd. | Hongshen (Beijing) Legal Consulting Co., Ltd.

Related documents: Terms of Service | Privacy Policy | Payment & Subscription Agreement | Developer Agreement | Account Deletion Guide