Model Provider Reference

This document summarizes the data handling, service level, and intellectual property indemnity commitments of the third-party AI model providers accessible through the Frontal platform. Anthropic (Claude) is the default active provider. All other providers listed are integrated with the platform and available for activation through your Account settings. You control which providers are active for your Account. This is a factual reference, not a legal commitment by Frontal. Provider terms may change. Confirm current terms directly with each provider before making procurement or compliance decisions. This document supplements the Subprocessors List, the AI and Machine Learning Disclosure, and the Data Processing Agreement.

1. PROVIDER SUMMARY TABLE

2. ZERO-RETENTION CONFIGURATIONS

2.1 When zero-retention is configured with a provider, inputs (prompts) and outputs (completions) are not retained by the provider after the API call completes. The provider does not store, log, or use the data for any purpose.

2.2 Zero-retention is available by default for API usage of Anthropic, OpenAI, Google Vertex AI, Microsoft Azure OpenAI, and Mistral AI. For ElevenLabs, check current terms.

2.3 For providers marked "configurable," zero-retention is the default for API usage but customers should verify their specific deployment configuration. For Google Vertex AI and Azure OpenAI, the customer's cloud configuration determines data handling.

2.4 Frontal contractually prohibits all model provider subprocessors from using Customer Data for training or improvement of their models. For providers that do not offer technical zero-retention configurations, Frontal relies on these contractual prohibitions. See our Data Disposal Policy, Section 4.3, for additional detail.

2.5 This table reflects API and enterprise-tier offerings. Consumer-tier products (e.g., ChatGPT, Claude.ai) are not used to process Customer Data through the Frontal platform.

3. PROVIDER SLA AVAILABILITY

3.1 Anthropic, OpenAI, Google Vertex AI, and Azure OpenAI offer enterprise-grade SLAs directly to their customers. These SLAs cover model inference availability. Frontal's Service Level Agreement covers the Frontal platform layer (AI Gateway routing, API endpoints, authentication, rate limiting). Individual model provider availability is the responsibility of the provider under their own SLA terms.

3.2 Where a customer has a direct enterprise relationship with a model provider (e.g., Azure OpenAI via the customer's own Azure subscription, or Anthropic via a direct enterprise agreement), the customer benefits from that provider's SLA directly. Where access is through Frontal's platform, Frontal's AI Gateway routes requests to alternative providers if one provider is unavailable.

3.3 For providers listed as "Contact provider," Frontal does not have visibility into or make representations about their SLA terms. Customers should contact the provider directly or consult the provider's published terms.

4. IP INDEMNITY FOR AI OUTPUTS

4.1 Anthropic, OpenAI, Google, and Microsoft offer IP indemnity for claims that AI-generated outputs infringe third-party copyright. These indemnities typically cover the customer's use of the provider's models for output generation but have terms, conditions, and exclusions that vary by provider. Key considerations:

• Anthropic Copyright Shield: Covers copyright infringement claims for outputs generated by Claude. Requires compliance with Anthropic's acceptable use policy.
• OpenAI Copyright Shield: Covers copyright infringement claims for outputs generated by OpenAI API models. Requires use of OpenAI's content filtering and compliance with usage policies.
• Google GCP IP Indemnity: Covers IP claims for customer use of Google Cloud services including Vertex AI. Subject to GCP terms.
• Microsoft Customer Copyright Commitment: Covers copyright claims for outputs from Azure OpenAI Service. Requires use of Microsoft's content filtering and guardrails.
• Mistral AI: Review Mistral's current commercial terms. Coverage may be more limited than the providers above.
• ElevenLabs: Review ElevenLabs' current commercial terms.

4.2 Frontal's own IP indemnity in the Terms of Service (Section 10) covers the Frontal platform itself. It does not cover outputs generated by third-party AI models. Customers who require IP indemnity for AI-generated outputs should verify coverage under the applicable model provider's terms.

4.3 Frontal does not guarantee that any provider's indemnity will apply in a given circumstance. Customers are responsible for understanding the scope, conditions, and exclusions of each provider's indemnity and for ensuring their use complies with the provider's terms.

5. HOW TO USE THIS DOCUMENT

5.1 Enterprise procurement and compliance teams: Use this table to identify which providers meet your organization's requirements for data handling, SLA, and IP protection. Configure your Frontal AI Gateway to route requests accordingly.

5.2 Developers and product teams: Use the zero-retention column to determine which providers are suitable for processing data subject to GDPR, CCPA, or internal data handling policies.

5.3 Legal teams: This document provides a summary. The definitive terms are the provider's own terms of service, DPA, and SLA. Links to current provider documentation are maintained in the Frontal Documentation at https://docs.frontal.dev/providers.

6. UPDATES

Provider terms change. This document is reviewed and updated quarterly, or upon material changes to provider terms that Frontal becomes aware of. The effective date reflects the most recent comprehensive review. For the most current information, consult the provider directly.

7. CONTACT

Frontal Labs, Inc. 131 Continental Drive, STE 305 Newark, DE 19713 United States of America Email: legal@frontal.dev