Acceptable Use Policy

This Acceptable Use Policy (“AUP”) is part of the SemanticGuard Terms of Service and applies to all use of the SemanticGuard service by you, anyone you authorize, and any application or end user that routes traffic through your account. Violations may result in suspension or termination of your account without notice and without refund.

1. Prohibited Content

You may not use SemanticGuard to generate, store, cache, or transmit:

• Child sexual abuse material (CSAM) or any content that sexually exploits minors. We will report suspected CSAM to NCMEC and applicable authorities as required by law.
• Content that incites or threatens violence, terrorism, or unlawful harm against any person or group.
• Content that infringes the intellectual property, publicity, or privacy rights of any third party.
• Malware, ransomware, exploits, or instructions to create them; phishing kits or social-engineering templates targeting third parties.
• Material that is defamatory, libelous, or knowingly false in a manner intended to harm a specific person or organization.
• Personally identifying information of third parties that you do not have a lawful basis to process, including biometric data and government identifiers.
• Content prohibited by the usage policy of the upstream LLM provider whose API you route through SemanticGuard (OpenAI, Anthropic, Google, Microsoft, Amazon, Mistral, and others). The upstream provider's policy applies to every request you send through their endpoint, regardless of routing.

2. Prohibited Activity

• Attempting to bypass, evade, or amplify rate limits or quotas of any upstream LLM provider, of SemanticGuard itself, or of any third-party service.
• Scraping, harvesting, or systematically extracting content from any third-party service in violation of that service's terms.
• Probing, scanning, or testing the security of SemanticGuard infrastructure or any upstream provider without prior written authorization. Coordinated vulnerability disclosure via security@semanticguard.dev is welcome and exempt.
• Using SemanticGuard to amplify or relay denial-of-service traffic against any target.
• Using SemanticGuard for cryptocurrency mining or any workload whose purpose is to inflate compute spend.
• Reverse-engineering, decompiling, or disassembling the SemanticGuard service, except to the extent applicable law permits despite this restriction.
• Misrepresenting your identity, your organization, or the nature of your application when registering an account or contacting support.
• Sharing, selling, or renting your SemanticGuard API key or account credentials to any third party not authorized under your subscription.

3. Prompt-Injection and Upstream-Safety Bypass

You may not use SemanticGuard to craft, store, or distribute prompts whose intent is to defeat the safety systems of an upstream LLM provider, including jailbreaks, system-prompt extraction attacks targeting a specific third-party application, or instructions designed to make an upstream model produce content the provider has classified as unsafe. Security research targeting your own systems is exempt; please coordinate with the upstream provider through their published vulnerability programs.

4. High-Risk Applications

SemanticGuard is not designed for use as the sole or primary decision-maker in any high-risk application, including medical diagnosis or treatment, legal advice, autonomous vehicle operation, critical infrastructure control, or any application where a wrong or stale response could cause physical injury, significant financial harm, or violation of fundamental rights. If you build such an application, you must layer human review, fresh-only routing, or another safeguard between SemanticGuard and the decision; you accept all risk of doing otherwise.

5. Reporting Abuse

If you believe content or activity on SemanticGuard violates this policy, report it to abuse@semanticguard.dev. For coordinated vulnerability disclosure, email security@semanticguard.dev.

6. Enforcement

We may investigate suspected violations using account metadata, audit logs, and (when Store Prompts is enabled on your account) the contents of stored prompts and responses. Confirmed violations may result in:

• Warning and required remediation;
• Suspension of the offending API key or account, with or without notice;
• Termination of your subscription without refund of any prepaid fees;
• Cooperation with law enforcement where required by law.

We reserve the right to take immediate action without prior notice for violations that pose imminent harm, that endanger the availability or security of the service for other customers, or that put our access to upstream LLM providers at risk.

7. Changes to This Policy

We may update this AUP from time to time. Material changes will be announced at least 30 days before they take effect, except where a shorter timeline is necessary to address an active threat or legal obligation. Continued use of the service after a change constitutes acceptance of the updated AUP.