What AI vendors commit to, in their own words.

Name: Anteroom AI vendor commitments
Creator: Anteroom
Keywords: AI vendor commitments, AI vendor commitment tracker, AI vendor commitment matrix, no-train default, IP indemnity, customer copyright commitment, copyright shield

Seven commercial commitments tracked across AI vendors: acceptable use policy, no-train default, output ownership, IP indemnity program, retention default, deprecation notice, and safety framework version. Every cell is anchored to a primary source URL. No interpretation. The product is structured citation.

Vendors

Commitments per vendor

5 / 105

Verified cells

These vendors appear in 20 tracked AI strategic partnerships. Each row is a primary-source-anchored architectural read.

Browse partnerships →

Showing 15 of 15 vendors

Anthropic

Claude

AUP: Usage Policy (effective September 15, 2025)
No-train default: Anthropic may not train models on Customer Content (Inputs + Outputs) under Commercial Terms
Output ownership: Customer retains all rights to Inputs; owns Outputs
IP indemnity program: Pending verification (Claude Enterprise terms)
Retention default: API: 30 days for safety. Enterprise: 0 days available.
Deprecation notice: At least 60 days notice before retirement for publicly released models
Safety framework: Responsible Scaling Policy v3.3 (effective May 26, 2026)

AWS Bedrock

Amazon Bedrock

AUP: AWS Acceptable Use Policy
No-train default: Customer data in Bedrock not used to train base foundation models
Output ownership: Customer owns Outputs
IP indemnity program: Uncapped IP indemnity for covered Bedrock services
Retention default: Not retained beyond inference unless customer enables logging
Deprecation notice: Ad-hoc per underlying model provider. Bedrock relays provider deprecation.
Safety framework: Relies on underlying model providers (Anthropic RSP, Meta, etc.)

Cohere

Command, Embed, Rerank

AUP: Cohere Usage Policy
No-train default: Customer prompts and outputs not used to train Cohere models by default
Output ownership: Customer owns Outputs
IP indemnity program: Pending verification (Cohere commercial terms)
Retention default: Pending verification (Cohere data processing terms)
Deprecation notice: Ad-hoc per model release
Safety framework: Cohere Responsible AI

Databricks

Mosaic AI, DBRX

AUP: Databricks Acceptable Use Policy
No-train default: Customer data in Mosaic AI Model Serving not used to train base foundation models
Output ownership: Customer owns Outputs
IP indemnity program: Indemnity available for Covered AI Services
Retention default: Not retained beyond inference unless customer enables logging
Deprecation notice: Standard Databricks product lifecycle policy
Safety framework: Databricks AI Security Framework (DASF)

Google Vertex AI / Gemini

Gemini, Vertex AI, Google DeepMind

AUP: Generative AI Prohibited Use Policy
No-train default: Vertex AI: customer data not used to train base models. Consumer Gemini: distinct policy.
Output ownership: Customer owns Outputs
IP indemnity program: Indemnified Generative AI Services (capped, named services)
Retention default: Vertex 30 days default; configurable lower
Deprecation notice: Ad-hoc per product. No standardized cross-product commitment.
Safety framework: Frontier Safety Framework (DeepMind)

Hugging Face

HF Hub, Inference Endpoints

AUP: Hugging Face Content Policy
No-train default: Hosted Inference Endpoints: customer data not used to train HF models. Hub uploads: model-by-model per uploader license.
Output ownership: Customer owns Outputs from Inference Endpoints; hosted model licenses vary per model
IP indemnity program: No standalone IP indemnity program for hosted services
Retention default: Pending verification
Deprecation notice: Per-model deprecation notices via Hub. No platform-wide commitment.
Safety framework: Hugging Face Responsible AI Posture

IBM

watsonx, Granite

AUP: IBM Acceptable Use Policy
No-train default: watsonx.ai: customer data not used to train Granite base models by default
Output ownership: Customer owns Outputs
IP indemnity program: IBM IP indemnity for Granite (Covered AI Services)
Retention default: Configurable per deployment
Deprecation notice: Standard IBM product lifecycle policy
Safety framework: IBM AI Ethics and Trustworthy AI