Terms of Use
Last updated: October 27, 2026
These Terms apply to the Anifold website and products and services that expressly refer to them, including safety environments and data, custom models and agents, capability evaluation, red teaming, safety alignment, post-training services, and the Safe Recursive Intelligence (SRI) platform. Please review these Terms before using the relevant services.
1. Scope and Agreements
Website content describes our research, products, and potential engagements. It is not a commitment to a particular delivery scope, performance level, price, or release date. Paid services, trials, model deliveries, data licenses, and training projects are governed by confirmed orders, service agreements, or other specific terms. More specific terms prevail for the matters they cover if they conflict with these Terms.
Services requiring acceptance of these Terms should present them appropriately and obtain valid acceptance. Publishing these Terms does not replace legally required notice, explanation, or consent procedures. You must have authority to act for any organization you represent. Individuals without independent contractual capacity must obtain the authorization required by law.
2. Use and Account Security
You may use the services for lawful business, research, training, and evaluation within the applicable agreement and license. Provide accurate necessary information, protect accounts, credentials, and API keys, and allow access only to authorized personnel. Notify us promptly of compromised credentials, unauthorized use, or security concerns and take reasonable protective steps.
You must not interfere with services, circumvent access or usage restrictions, access another party's accounts or data, or copy, resell, or distribute restricted resources without authorization. Modification, sublicensing, and redistribution of software, models, or data must comply with the relevant license. These Terms do not restrict activities expressly permitted by law that cannot lawfully be excluded by contract.
3. Authorized Security Research
Safety environments, attack knowledge, malicious examples, and red-team tools may have dual uses. Use them only with systems and data you own or are expressly authorized to test, and within that authorization. Establish the permitted assets, methods, time windows, and isolation measures before testing. Access to research or evaluation resources does not authorize testing third-party targets.
Do not use the services for unauthorized intrusion, data theft, malware distribution, fraud, or other unlawful conduct. Adversarial evaluation and harmful-content analysis require lawful authorization and appropriate controls. Test materials must not be used to infringe rights or damage real systems. We may apply restrictions proportionate to the risk and violation involved.
4. Customer Content and Processing Authorization
You or the relevant rights holders retain existing rights in submitted data, prompts, files, models, configurations, and other materials. You are responsible for a lawful basis and necessary permissions to provide and use them, including compliance with privacy, confidentiality, intellectual property, and other applicable requirements.
You authorize us to process customer content only as necessary for the agreed services, including running environments, executing rollouts, generating evaluations, and performing authorized training. Our Privacy Policy and applicable data processing agreements address data protection. Without express authorization or a separate written agreement, we do not use non-public customer content to train general-purpose models for other customers or supply it in external datasets.
5. Models, Data, and Deliverables
The relevant agreement specifies delivery formats, acceptance criteria, ownership, and usage rights for custom model weights, adapters, agent systems, training data, code, evaluation reports, and other deliverables. Paying for services or obtaining access does not by itself transfer all intellectual property or grant redistribution rights in third-party materials.
We and our licensors retain rights in pre-existing software, platforms, tools, methods, trademarks, documentation, and other background technology. Open-source models, software, and third-party data remain subject to their respective licenses. Our agreements cannot grant rights beyond those permitted by the relevant rights holders. Use of customer deliverables must respect applicable agreements, existing customer rights, and confidentiality obligations.
6. Outputs, Evaluations, and Oversight
Model outputs, automated scores, red-team findings, and evaluation reports may contain errors, omissions, or conclusions that do not hold in deployment. Results depend on test data, scope, environment configuration, and model version. They do not establish safety in every context or constitute regulatory approval or compliance certification unless the relevant credentials and scope are expressly identified.
Verify outputs for your intended use and apply appropriate human review and technical controls to important decisions, real-system operations, and high-risk applications. Specific performance commitments, safety metrics, and service levels apply only as expressly agreed. These Terms do not excuse obligations we owe under applicable law or an agreement.
7. Fees, Third-Party Resources, and Confidentiality
Orders or service agreements specify fees, billing, compute and usage limits, payment, renewal, and refunds. Browsing the website alone does not incur service charges. Use of paid third-party models, cloud resources, or other resources should be addressed in the applicable service arrangements.
Non-public technical, business, and customer information received during an engagement is protected under applicable confidentiality or service terms. Public website content is not inherently confidential and does not grant permission to use Anifold's or a customer's name for publicity.
8. Changes, Suspension, and Termination
We may change website or product features. Material changes, discontinuation, and migration for contracted services are governed by the relevant agreement and law. Suspected unlawful, unauthorized, security-threatening, or materially breaching use may be restricted or suspended to the extent necessary and proportionate. Where feasible and legally permitted, we provide reasons and an opportunity to remedy the issue.
Data export, retention, deletion, and delivery after termination follow applicable agreements, our Privacy Policy, and legal requirements. Termination does not affect accrued lawful rights and obligations or provisions intended by their nature to survive, including confidentiality, intellectual property, and dispute-resolution provisions.
9. Liability and Events Beyond Control
Our liability is determined by applicable law and the relevant agreement. Any limitation or exclusion is interpreted and applied only to the extent legally permitted. It does not exclude liabilities that cannot lawfully be excluded, including applicable liability for personal injury and property loss caused by intentional misconduct or gross negligence.
If force majeure or another event affects performance, the parties must give legally required notice, take reasonable steps to mitigate loss, and determine responsibility under applicable law and agreements in light of the actual impact. Using third-party services does not automatically relieve either party of its legal obligations.
10. Applicable Law, Disputes, and Updates
Valid provisions in the relevant agreement govern applicable law and dispute resolution. Without such provisions, applicable conflict-of-laws, substantive-law, and jurisdiction rules determine those matters. The parties should first seek to resolve disputes through good-faith discussions, without limiting lawful remedies or non-waivable rights.
We may update these Terms and their date. Material changes affecting existing services require notice and acceptance where required by law or agreement. Updates do not automatically alter signed agreements or accrued rights retroactively. The Chinese and English versions are intended to convey the same meaning. Language precedence for a transaction is determined by its agreement, and neither version limits mandatory legal protections.
11. Contact Us
For questions about these Terms, authorization, purchasing services, or rights, contact [email protected]. Contracted customers may also use the channels specified in their service agreement.