These terms govern use of Forge, the enterprise AI platform operated by Phantom Forge Labs. Forge composes integrated platform capabilities — Forge Gate, Forge Portfolio, Forge Passport, and Forge Studio — in one organization.
Forge provides an organization-scoped web application for Phantom Forge Labs' customer-owned delivery journey — AI releases, evidence exports, review, and related APIs through the Forge platform — so you retain control of your data, IP, and brand. Features depend on your deployment configuration (Supabase, LLM providers, Stripe, email).
You are responsible for activity under your accounts, for maintaining credential secrecy, and for ensuring invited users comply with these terms.
You retain ownership of data you submit. The processor role for personal data, subprocessors, and transfers is described in the Data Processing Agreement (see /legal/dpa).
Forge supports mixed contract tiers in one organization. On Build tiers, production weights, corpora, and customer-cloud inference remain under your custody as sold. When Build cloud delivery is enabled, Forge may provision Inference API and/or Managed agents proxies that invoke models or agents in your connected cloud account (for example AWS Bedrock, GCP Vertex, or Azure OpenAI via cross-account credentials); compute and model spend stay in your tenancy. On the AI SaaS tier, when your order includes Forge-operated delivery, Phantom Forge Labs may host Inference API and/or Managed agents behind stable Forge aliases; the cloud substrate stays under PFL control and is not a customer-chosen pick list in the product UI. Build Managed agents currently require AWS AgentCore where that option is enabled.
If your workspace enables Forge model learning for AI SaaS, Phantom Forge Labs may use scrubbed product signal from AI SaaS releases (for example intent previews, option accepts, eval outcomes, and managed-inference metadata) to improve Forge-operated models offline and to promote improved backends behind the same customer-facing alias. Build-tier releases are excluded from that learning path by default. Learning does not grant PFL ownership of your customer-facing product, brand, or Build-tier IP.
No unlawful content, no attempts to break tenant isolation or probe other organizations, and no abuse of APIs or rate limits. AI outputs are assistive; you remain responsible for production decisions.
The service is provided “as is” to the maximum extent permitted by law. Fitness for a particular regulatory regime (for example FedRAMP) requires a separate written agreement and architecture review.
Material changes to these terms will be communicated with reasonable notice per your commercial agreement.