Legal Documents

1. Acceptance of Terms

By accessing or using the ModelSwitch AI API, platform, or any associated services (collectively, "Services"), you agree to be bound by these Terms of Service. If you are acting on behalf of a company or other legal entity, you represent that you have authority to bind such entity to these terms.

2. Our Services

ModelSwitch AI Inc. ("ModelSwitch", "we", "us") provides an LLM orchestration infrastructure layer that routes, proxies, and optimizes requests to third-party AI model providers. We do not operate the underlying language models. Your use of any AI-generated output is subject to the terms of the respective downstream provider.

3. API Access & Authentication

API keys issued to you are credentials that grant access to your account. You are responsible for maintaining the confidentiality of your API keys and must not share them with third parties or embed them in publicly accessible code repositories. Immediately rotate any key you believe may have been compromised.

4. Acceptable Use Policy

You may not use the Services to generate unlawful content, conduct mass surveillance, create disinformation at scale, or process data in violation of applicable law. We reserve the right to suspend accounts upon detection of policy violations without prior notice.

5. Open Source License

The core ModelSwitch routing engine is released under the MIT License. You may self-host, modify, and distribute it in accordance with that license. Managed cloud features, enterprise tooling, and the ModelSwitch brand are not covered by the open-source license.

6. Limitation of Liability

TO THE MAXIMUM EXTENT PERMITTED BY LAW, MODELSWITCH'S TOTAL AGGREGATE LIABILITY FOR ANY CLAIMS ARISING UNDER OR RELATED TO THESE TERMS WILL NOT EXCEED THE GREATER OF (A) THE AMOUNT YOU PAID TO MODELSWITCH IN THE 12 MONTHS PRIOR TO THE CLAIM, OR (B) USD $100. MODELSWITCH IS NOT LIABLE FOR ANY INDIRECT, INCIDENTAL, SPECIAL, CONSEQUENTIAL, OR EXEMPLARY DAMAGES.

7. Governing Law

These Terms are governed by the laws of the State of Delaware, USA, without regard to conflict of law principles. Disputes shall be resolved exclusively in the state or federal courts located in Delaware.

8. Changes to Terms

We may update these Terms at any time. Material changes will be announced 30 days in advance via the email on file. Continued use of the Services after the effective date constitutes acceptance of the revised Terms.

Data We Collect

We collect: (a) Account data — name, email, company, billing address; (b) Usage data — API request logs including timestamps, model IDs, token counts, routing decisions, and latency; (c) Technical data — IP addresses, browser type, and device identifiers for security purposes. We do not store the content of your prompts or completions on our persistent storage beyond the 72-hour rolling window required for abuse detection.

How We Use Data

Data is used to: provide and improve the Services, authenticate API requests, generate billing statements, detect and prevent abuse, and comply with legal obligations. We do not sell your data to third parties.

Data Residency & Transfers

By default, request metadata is stored in our US-East (Virginia) region. Enterprise customers may elect EU (Frankfurt) or APAC (Singapore) residency. Transfers from the EEA to the US are made under Standard Contractual Clauses (SCCs) under GDPR Article 46.

Your Rights (GDPR/CCPA)

Depending on your jurisdiction, you have the right to access, correct, delete, restrict processing of, and port your personal data. To exercise these rights, contact legal@modelswitchai.tech. We will respond within 30 calendar days.

Cookies

Our marketing site uses essential cookies for session management and anonymized analytics (PostHog). No advertising cookies. API traffic is cookie-free. Manage preferences via the Cookie Policy section below.

Uptime Commitments

Latency SLAs

For Enterprise managed deployments, the P95 proxy overhead shall not exceed 100ms for requests under 8,192 input tokens. Streaming requests have a separate time-to-first-token SLA of P95 <150ms, agreed per contract.

Downtime Definition

Downtime is defined as any 5-minute window in which error rates exceed 10% on the /api/v1/chat/completions endpoint, excluding scheduled maintenance with 48h notice, third-party provider outages, and force majeure.

Credits

SLA credits for Enterprise customers are detailed in your subscription agreement. To claim a credit, contact legal@modelswitchai.tech within 30 days of the incident with supporting data.