Definitions
A Subprocessor is a third party engaged by AI Reserve to process Personal Data on our clients' behalf, acting only on our documented instructions.
Under our Data Processing Addendum, a provider that processes data for its own purposes — for example, a model provider whose terms permit it to train or fine-tune on inputs — is an independent controller, not a subprocessor. Such providers are identified separately in the Independent controllers section below. Requests are only sent to those providers when a user or client explicitly selects one of their models.
Per-provider detail on training, retention, and processing region is on the Provider Data Handling page.
Infrastructure & platform subprocessors
These providers support the operation of the AI Reserve platform itself and may process Personal Data in the course of providing their services.
| Subprocessor | Function | Jurisdiction |
|---|---|---|
| Google Cloud Platform | Cloud hosting, compute, databases, object storage, secret management, and logging for all AI Reserve services | United States |
| Google Firebase Authentication | Identity and sign-in for the AI Reserve portal (web application) | United States |
| Auth0 by Okta | OIDC identity brokering for keyless desktop-application sign-in (e.g. Claude Desktop) | United States |
| Resend | Transactional email delivery for the AI Reserve portal (onboarding invitations, password resets, account and allocation notifications, and internal operational reports). Receives recipient names, email addresses, and notification content — never prompt or response content | United States |
| IPinfo | IP geolocation and anonymizer screening for the machine API. Receives request IP addresses only — never request content | United States |
| GDPRLocal Ltd. / Instant EU GDPR Representative Ltd | Article 27 EU/UK representative (GDPR Article 27 and UK GDPR Article 27). Receives privacy requests submitted through its portal on our behalf — requester identity and request content only; never prompt or response content | Ireland / United Kingdom |
| Slack Technologies (Salesforce) | Slack Connect shared channels for customer communications, created from the AI Reserve admin console. For customer organizations that enable a Slack Connect channel, Slack receives the customer admin's email address (used to send the Slack Connect channel invite), a shared channel named for the customer organization, and the messages participants post in that shared channel — including, where the organization uses the @aireserve mention agent, prompts addressed to the agent and the model completions the agent posts back. Slack also receives operational alert payloads that may reference customer organization names. Content sent through the API directly does not transit Slack | United States |
| Microsoft Corporation | Microsoft Teams customer channels and Microsoft Entra ID sign-in. Microsoft Teams: for customer organizations that choose a Microsoft Teams channel, AI Reserve staff communicate with the customer's contacts in a private channel in AI Reserve's Microsoft 365 tenant, to which those contacts are invited as guests. Microsoft receives the customer contacts' names and work email addresses (sent at onboarding, when the Microsoft Teams channel option is selected, to create the channel and send the guest invitations), a channel named for the customer organization, and the messages and files participants choose to share in that channel — business communications only. No prompts, responses, or other content sent through the platform transits Teams, and the channel has no agent or bot. Microsoft Entra ID: for organizations that deploy the Microsoft Office integration (the Claude for Microsoft 365 add-in) or use Connect with Microsoft, Microsoft Entra ID issues the sign-in tokens the platform verifies; those tokens carry the user's directory (tenant) id, object id, and sign-in email address. The platform sends no prompt, response, or other customer content to Microsoft on this path. Processed under the Microsoft Products and Services Data Protection Addendum. Listed September 16, 2026 | United States |
| Daytona Platforms Inc. | Isolated, single-use workspaces for the customer Slack agent (@aireserve); no code activation keyword is required. Each admitted request runs its agent harness inside one private Daytona sandbox, whether it answers a question, executes scratch code, or works with an authorized repository. GitHub connection is optional and needed only for repository tools. The sandbox receives the request text and bounded surrounding Slack thread context, an execution-scoped credential for AI Reserve's inference gateway, and, only if the agent checks out an authorized repository, that repository's contents and a short-lived read-only single-repository token; its network access is limited to that gateway, GitHub, and the npm, Yarn, and PyPI package registries. Model inference is served through the gateway by the organization's selected model provider, never by Daytona. The sandbox is created ephemeral and deleted at the end of the run, with a provider-enforced time limit as backstop; answers return to the Slack thread, and repository changes return as a pull request in the customer's own GitHub repository. A private final-result artifact, including the answer and any proposed changes, is held in AI Reserve's Google Cloud storage for seven days so delivery/publication can be recovered. No Daytona management credential, GitHub write token, or Google Cloud credential enters the sandbox. Daytona's retention of sandbox data after deletion: Not published — pending verification. Processed under the Daytona Data Processing Agreement. Listed September 17, 2026, ahead of first production use |
United States (New York company; sandboxes run in Daytona's United States region) |
| Runpod Inc. | Customer-configured GPU compute — on-demand GPU pods and persistent network volumes an organization provisions from the AI Reserve console. Enabled in our production and UAT environments for organizations whose administrator has turned GPU compute on. RunPod receives the configuration an organization supplies for each pod — the container image reference, environment variables and start-up settings, the GPU type and data center the organization selects, the SSH public keys used to reach the pod, and, for private images, the container-registry credentials the organization enters — and hosts the running containers and the network volumes the organization creates, resizes, or deletes, including whatever the organization stores on them. Pods run on RunPod Secure Cloud; RunPod's DPA states hosting on servers throughout the world, including the United States, and AI Reserve has not yet restricted the data centers available to our account. Prompts and responses sent through the AI Reserve gateway do not transit RunPod. A pod's container storage is discarded when the pod is terminated; network volumes persist until the organization deletes them. RunPod's retention of container or volume data after deletion: Not published — pending verification. Processed under the Runpod Data Processing Agreement. Listed September 17, 2026 | United States (company); pods and volumes run in the RunPod Secure Cloud data center the organization selects, which may be outside the United States |
| Apollo.io (ZenLeads Inc.) | Business-contact and company enrichment for the optional chat personas (the contact-search and company-team tools) and for the company profile page's team-member and technology-stack sync. Not enabled — listed ahead of enablement: the integration exists in code, but the only deployed service that mounts the credential has no Apollo call site, and the web app — where every deployed Apollo call site lives — does not mount it, so no data has been sent to Apollo. If it were enabled, Apollo would receive the identifiers a user asks a persona to look up — a person's name and company domain, a LinkedIn profile URL, or an email address (information about third parties, not about the organization's own users) — and, on company profile pages, the company's domain; it would return that person's business and personal email addresses, phone numbers, title, and location, or the company's team members and technology stack. No other prompt or response content would transit Apollo. Apollo's DPA treats Apollo as our processor when matching the identifiers we send, and as an independent controller of its own contact database, which Apollo's privacy policy says customer-submitted data may be used to enrich; whether lookups made through the API for our account would feed that database: Not published — pending verification. Whether to enable or remove the integration is under review with counsel. Terms: Apollo Data Processing Addendum, privacy policy, and subprocessor list at trust.apollo.io. Listed September 17, 2026, ahead of enablement | United States (EU-U.S. Data Privacy Framework participant; Standard Contractual Clauses as fallback per its DPA) |
| Exa Labs, Inc. (Exa) | Web search for the portal chat assistant. When a chat request calls for current information from the web, the model generates a search query derived from the user's prompt and the platform sends that query text to Exa's search API, which returns public web results (page titles, addresses, and excerpts) that the response cites. Exa receives the search query text — which can contain terms the user supplied — and nothing else: no other conversation content, no documents, no account data, and no user or organization identifiers. Content sent through the API directly does not transit Exa. Exa's retention of query text for our account: Not published — pending verification (Exa offers Zero Data Retention on its Enterprise plans, enabled per team; not yet verified for our account). Exa's privacy policy states that query data submitted to its search products is used to improve them, including model training, and that content Exa processes as a processor for business customers is governed by their agreements instead; its Data Processing Addendum limits processing of personal data to our documented instructions. Whether any query text sent through the API for our account is used to improve Exa's services: Not published — pending verification; under review with counsel. Processed under the Exa Data Processing Addendum; privacy policy. Listed September 28, 2026 | United States |
| Composio (Sampark, Inc. d/b/a Composio) | Connector platform for the portal's optional app connectors. A user may connect their own accounts on third-party applications — Gmail, Google Calendar, Google Docs, Google Drive, Outlook, OneDrive, Microsoft Teams, Slack, Zoom, Box, Dropbox, GitHub, GitLab, Jira, Confluence, Linear, Asana, ClickUp, Monday, Notion, Airtable, HubSpot, Salesforce, Intercom, Zendesk, QuickBooks, Square, and Stripe — from the portal's Tools page. Composio brokers the sign-in and authorization with that application, holds the resulting authorization (the OAuth grant and tokens, encrypted at rest per its Data Processing Addendum) on the user's behalf, and executes connector tool calls when the user attaches a connected application to a chat conversation and the chat assistant uses one of its tools. For each such tool call, Composio receives the arguments the model generates (which can include content from the user's prompt), relays the call to the connected application using the user's own authorization — read actions and, where the organization's deployment enables them, write actions such as sending a message or creating a calendar event — and returns the application's response (for example messages, calendar entries, files, records, or issues from the user's own account) to the platform. The platform identifies the user to Composio by an opaque platform user id, not by name or email address. Each connection is initiated by the user, is scoped to that user's own accounts, and can be revoked by the user from the Tools page at any time (revoking deletes the connected account and its tokens at Composio). No other conversation content, documents, or account data is sent to Composio; connector tools are present only in a conversation to which the user has attached a connected application; and content sent through the API directly does not transit Composio. Composio's retention of tool-call request and response payloads: Not published — pending verification (its Data Processing Addendum states that request and response payloads are logged where logging is enabled; the setting for our account is not yet verified). Processed under the Composio Data Processing Addendum (entered into by acceptance in the Composio dashboard — acceptance for our account: pending verification); privacy policy; subprocessor list at trust.composio.dev. Listed September 28, 2026 | United States (Delaware company, San Francisco) |
Model providers acting as subprocessors
When a user or client selects a model, the request content is processed by the model's
provider. The following providers process request data only to deliver inference, do
not train on API inputs under their current enterprise terms, and act as subprocessors.
A request is sent to the provider serving the model selected — no provider receives
traffic for models it does not serve. If that provider fails or is saturated, the
platform may retry the request through a disclosed failover chain containing only
deployments of the identical model weights hosted by another
subprocessor on this page (for example, Anthropic Claude models on AWS Bedrock).
The platform default never substitutes a different model; cross-model entries exist
only where a client administrator has explicitly configured them for their own
organization, and models with no second same-weights host fail with the provider's
real error. Every response served by a fallback is disclosed per response via the
x-aireserve-served-model header. Failover never sends a request to a
provider not listed on this page, and never to an
independent controller unless a client
administrator has explicitly configured that for their own organization. See
provider failover on the Data Handling page.
| Subprocessor | Function | Jurisdiction |
|---|---|---|
| OpenAI, L.L.C. | Model inference (GPT model family; embeddings; image generation) | United States |
| Anthropic, PBC | Model inference (Claude model family, direct API) | United States |
| Amazon Web Services (Bedrock) | Model inference for -bedrock routed models (Anthropic Claude, Meta Llama, Mistral, DeepSeek serverless) — model publishers do not receive request data |
United States |
| Google Cloud (Vertex AI) | Model inference for -vertex routed models (Anthropic Claude partner models in Vertex AI Model Garden) — the model publisher does not receive request data |
United States (us-east5) |
| Google (Gemini API) | Model inference (Gemini model family; image generation), paid tier. Also indexes documents saved to the web document library for chat search: each saved document is uploaded into a Gemini File Search store scoped to your organization, where Google chunks, embeds, and stores its content with no expiry; the platform deletes the indexed copy when the document is deleted and confirms the deletion against the store, a daily sweep removes any copy a deletion could not complete, and the organization's store is deleted when the organization is closed (see data handling, including the one-time clean-up of copies an earlier defect left behind) | United States (processing may occur globally — see data handling) |
| xAI Corp. | Model inference (Grok model family); text-to-speech and customer-created custom voices — organization-gated, off by default for every organization; when enabled, uploaded reference audio is processed by xAI and the resulting voice model is stored by xAI until deleted (see data handling) | United States |
| Perplexity AI, Inc. | Model inference with web search grounding (Sonar model family) | United States |
| Together Computer, Inc. (Together AI) | Hosted inference of open-weight models (e.g. Qwen, DeepSeek V4 Pro, Kimi K2.7, MiniMax, GPT-OSS-20B). For organizations that use the Fine Tuning marketplace (since September 24, 2026): receives the organization's uploaded training dataset (streamed directly from the browser or the streaming relay, never stored by AI Reserve) and hosts that dataset and the resulting fine-tuned weights under AI Reserve's account until the organization deletes them; the fine-tuned model is served only to that organization — see data handling | United States |
| Fireworks AI, Inc. | Hosted inference of open-weight models (e.g. GLM, Qwen, Kimi K2.6, DeepSeek V4 Flash, GPT-OSS-120B). Fine Tuning marketplace path implemented (LoRA, provider-signed browser uploads, dedicated deployments): Not enabled — the fine-tuning account binding is not configured in any environment, so no training dataset has been sent; once configured, Fireworks hosts the organization's dataset and fine-tuned weights under AI Reserve's account until the organization deletes them — see data handling | United States |
| DeepInfra, Inc. | Hosted inference of open-weight and partner models (NVIDIA Nemotron 3 family, DeepSeek V4 Flash, Qwen 3.8 Max) | United States (Delaware corporation, HQ Palo Alto, CA; SOC 2 and ISO 27001 certified; US data centers plus one Canadian site — Toronto) |
| Novita AI | Hosted inference of open-weight and partner models (DeepSeek V4.1 Flash, DeepSeek V4 Flash until October 9, 2026, Qwen 3.8 Max) | United States (San Francisco HQ; SOC 2 Type II audited; processing regions vary — see data handling) |
| Baseten, Inc. | Hosted inference of open-weight models (DeepSeek V4 Flash family, GPT-OSS 120B). Fine Tuning marketplace path implemented (Baseten Loops, LoRA, SDK training worker reading the organization's own bucket): Not enabled — the training worker is not provisioned in any environment, so no training data has been sent; once enabled, Baseten receives bounded training batches and hosts the resulting checkpoint under AI Reserve's workspace until the organization deletes it — see data handling | United States (San Francisco HQ; SOC 2 Type II audited and HIPAA compliant per trust center; serving region for our account — see data handling) |
| Alibaba Cloud (Model Studio) | Model inference (first-party Qwen models — Qwen 3.8 Max, direct Model Studio API) | United States (Virginia) for Qwen chat models (dashscope-us endpoint), except qwen-3-coder-next, which is served from Singapore (dashscope-intl endpoint — the only region offering that model, effective September 1, 2026); Singapore for Wan video generation and Qwen image generation (qwen-image-3.0 / qwen-image-3.0-pro, dashscope-intl endpoint); request data stored in the selected region. The operator is a PRC-headquartered group, so the platform labels these models China-based and serves them only to organizations that have not opted out of PRC-affiliated models — see data handling |
| MuleRouter (mulerouter.ai — CarrotHub / SmartStudio ecosystem) | Hosted video generation for the Wan video model families (Wan 3.0 / W3.0, and Wan 2.7 image-to-video), served without upstream content moderation. Requests carry the video prompt and any customer-supplied media inputs — source/keyframe images (a source image is required by the Wan 2.7 image-to-video model), reference images, and reference/driving audio or video, passed by URL or inline — only for organizations that are expressly allowlisted for adult-content models, whose administrator has enabled adult-content serving, and whose caller has accepted the third-party model policy; these models are never publicly listed. MuleRouter's documentation states generated outputs belong entirely to the user. No training use, and tasks are deletable on request — confirmed in writing by the Alibaba Cloud International partner team (August 26, 2026; written vendor confirmation, not yet published terms); no published fixed deletion window — see data handling | Singapore — confirmed in writing by the Alibaba Cloud International partner team (August 25, 2026); the operator is an Alibaba Cloud–partnered entity serving PRC-developed (Wan) models; the platform labels these models China-based and serves them only to organizations that have not opted out of PRC-affiliated models |
| WaveSpeedAI PTE. LTD. (wavespeed.ai) | Hosted video extension and prompt-driven image editing for the Wan 2.7 family (video extension continues an existing customer-supplied video clip, guided by a prompt and optional driving audio; image editing makes targeted changes to one to nine customer-supplied reference images, guided by a prompt; all media passed by URL). Video extension serving since September 1, 2026 (launch gate enabled by a reviewed change); image editing added September 2026 on the same account and launch gate. Requests are served only for organizations that are expressly allowlisted for adult-content models, whose administrator has enabled adult-content serving and — while the retention terms below are pending — the unverified-retention consent, and whose caller has accepted the third-party model policy; these models are never publicly listed. Training, retention, and deletion terms pending verification — see data handling | Singapore-incorporated operator (WaveSpeedAI PTE. LTD., per its published privacy policy); serving region and upstream affiliations pending verification — conservatively labeled China-affiliated until verified, and served only to organizations that have not opted out of PRC-affiliated models |
| Video Rebirth Limited (BACH by Video Rebirth, bach.art) PRC and data-retention consents required |
Hosted video generation with the BACH bach-1.0 engine: text-to-video, image-to-video, and reference-to-video (bach-1.0-video, bach-1.0-i2v, bach-1.0-reference). Receives the generation prompt and any customer-supplied first-frame, last-frame, or reference images by public URL, and hosts the generated video at a result link. Listed September 29, 2026 ahead of deployment, pricing, and live verification; the routes stay unavailable until credentials and pricing are recorded, and are then served only to organizations whose administrator has enabled BOTH the PRC-affiliated-models and the unverified-retention consents. Training, retention, and deletion terms for the API are not published pending verification — see data handling |
Singapore headquarters per the operator's own press releases and website terms (Singapore governing law); Hong Kong / Singapore / United States offices per its contact page; legal entity name ("Video Rebirth Limited" in its corporate policies, "Video Rebirth Private Limited" in bach.art's structured data), API processing region, and upstream affiliations pending verification — conservatively labeled China-affiliated until verified, and served only to organizations that have not opted out of PRC-affiliated models |
| monid.ai | Data-endpoint marketplace behind /v1/monid (discover, inspect, run) — brokers third-party data endpoints (web scrapers, contact/company enrichment, search) that AI Reserve API callers execute with their own AI Reserve key. Receives the natural-language discovery query and the endpoint inputs the caller supplies for a run, and forwards run inputs to the third-party provider the caller selected (named in every response). Never receives chat, completion, or document content. Training and retention terms not published pending verification — see data handling |
Not published pending verification |
| TypeSafe AI | Automatic model selection when the Prompt Router is active: receives up to 4,000 characters from the beginning and end of the latest user message, organization routing instructions/preferences, model catalog facts, and structural request parameters; returns typed model choices. This platform-paid routing step is bypassed by concrete model selections and is separate from the following classification opt-in. Per-request business-function classification (optional, separate administrator opt-in per organization). Receives a bounded excerpt of each classified request — the latest user message (at most 2,000 characters), the head of the system prompt (at most 600 characters), declared tool names — plus the API-key and team names, and returns a typed category with probabilities; never generates text, never receives full conversations, completions, or documents. The excerpt is processed in memory and never stored by AI Reserve. Training posture: customer data does not influence model training per TypeSafe's published model documentation; retention terms not published pending verification — see data handling | Not published pending verification |
| Nebius B.V. (Token Factory) | Fine Tuning marketplace path implemented (LoRA and full fine-tuning through the Token Factory Files and fine-tuning APIs; checkpoint export): Not enabled — no credential is provisioned in any environment, so no data has been sent. Once enabled, Nebius receives the organization's training dataset (streamed through the relay, never stored by AI Reserve) and hosts that dataset and the trained checkpoints under AI Reserve's account until the organization deletes them. Training-use and retention terms for our account: Not published — pending verification — see data handling | Netherlands-headquartered group; processing region for our account: pending verification |
| Thinking Machines Lab, Inc. (Tinker) | Fine Tuning marketplace path implemented (LoRA training through the Tinker SDK from AI Reserve's training worker, which reads the organization's own bucket; low-volume preview sampling endpoint): Not enabled — no credential is provisioned and the training worker is not deployed in any environment, so no data has been sent. Once enabled, Thinking Machines receives bounded, tokenized training batches and hosts the resulting adapter weights under AI Reserve's account until the organization deletes them. Training-use and retention terms for our account: pending verification — see data handling | United States; processing region for our account: pending verification |
| Microsoft Corporation (Azure OpenAI) | Fine Tuning marketplace path implemented (Azure OpenAI fine-tuning jobs and deployments in an AI Reserve Azure subscription): Not enabled — no Azure OpenAI resource is provisioned for fine-tuning in any environment, so no data has been sent. Once enabled, Microsoft receives the organization's training dataset (streamed through the relay, never stored by AI Reserve) and hosts that dataset and the fine-tuned deployment under AI Reserve's subscription until the organization deletes them, under the Microsoft Products and Services Data Protection Addendum — see data handling | Region of the Azure OpenAI resource AI Reserve provisions: pending verification |
| Mistral AI (La Plateforme) | Model inference (first-party Mistral models — mistral-medium-3.5, codestral-2508, ministral-14b — direct La Plateforme API) |
European Union — France-headquartered provider; La Plateforme runs on EU-based infrastructure by default; serving region for our account pending verification — see data handling |
| OpenRouter, Inc. | Routing layer for certain open-weight model routes (Meta Llama, Mistral). OpenRouter itself does not store request content by default; the downstream serving endpoint's policy applies — per-route endpoints pending verification | United States (routing); downstream endpoint varies |
| fal — Features & Labels, Inc. (fal.ai) | Hosted video and image generation | United States |
| GMI Cloud | Hosted inference of open-weight serving legs (DeepSeek V4 family, GLM 5.2/5.3/5.3 Flash, Kimi K2.6/K2.7 Code — api.gmi-serving.com). No API-content training or retention terms published pending verification — GMI Cloud-served models are locked for every organization until its administrator enables the data-retention consent gate in the AI Reserve console; see data handling |
United States (San Jose, CA company); serving region for our account pending verification |
| Tensormesh (serverless inference) | Model inference (open-weight serving legs, serverless.tensormesh.ai). Retention and training-use terms pending verification — Tensormesh-served models are locked for every organization until its administrator enables the data-retention consent gate in the AI Reserve console; see data handling |
United States pending verification |
| ByteDance (BytePlus ModelArk) | Hosted image (Seedream) and video (Seedance) generation via the international ModelArk API. PRC-headquartered operator — models are labeled China-based and served only to organizations that have enabled PRC-hosted models; terms pending verification — see data handling | Singapore (BytePlus international, ap-southeast); operator headquartered in the PRC |
| StreamLake (Kuaishou "Vanchin" international platform) | Model inference (open-weight -streamlake serving legs, vanchin.streamlake.ai). StreamLake's international ToS (§4.6) license it to use inputs and outputs for model training with no published opt-out, and its parent Kuaishou is PRC-headquartered — StreamLake-served models are locked for every organization until its administrator enables both the PRC-hosted-models opt-in and the training consent gate in the AI Reserve console; see data handling |
Singapore (stated data-storage location); operator's parent headquartered in the PRC; inference hosting location unstated |
| Hugging Face, Inc. (Inference Endpoints) | Dedicated GPU inference endpoints that enterprise administrators deploy one-click from the AI Reserve console (open-weight models, e.g. Meta Llama, Mistral, Qwen, Phi, Gemma, BGE embeddings). Endpoints are provisioned under AI Reserve's Hugging Face organization as token-protected, single-tenant deployments and process the prompts and completions sent to them; they receive traffic only for the organization that deployed them, are billed hourly at Hugging Face list price from that organization's wallet, and are separate from gateway chat routing — see data handling | United States (New York HQ; endpoints in the current catalog provision on AWS us-east-1) |
Model providers acting as independent controllers
The following model providers' current terms permit them to process request data
for their own purposes (including using inputs to train or improve
their services). Under our DPA they are therefore independent controllers, not
subprocessors, and are identified as such. Requests reach these providers
only when the organization's administrator has enabled the training
consent gate in the AI Reserve console and a user or client explicitly
selects one of their directly-served models — as the requested model, or as an entry
the client's own administrator explicitly configured in that organization's fallback
chain (chain entries hosted in the PRC are excluded for organizations that have opted
out of PRC-hosted models, and every fallback serve is disclosed via the
x-aireserve-served-model header). Absent that consent, these providers'
models are locked for the organization — greyed out in the catalog and refused at the
gateway. Open-weight DeepSeek, Moonshot, Qwen,
and NVIDIA Nemotron models served by US aggregators (Fireworks, Together, DeepInfra,
Novita) or AWS Bedrock do not send data to these companies — nor, for the
Qwen serving variants, to Alibaba Cloud.
| Provider | Function | Jurisdiction | Why independent controller |
|---|---|---|---|
| Higgsfield Inc. Training and data-retention consents required |
Wan 3.0 text-to-video, image-to-video, and reference-to-video (wan3.0-video-higgsfield, wan3.0-i2v-higgsfield, wan3.0-reference-higgsfield). Receives generation prompts and caller-supplied image, video, and audio references by public URL and hosts generated videos. Listed September 22, 2026 ahead of deployment and live verification; all three routes require both existing organization consents. See data handling. |
United States (San Francisco headquarters); serving location and downstream AI/compute providers for our API account pending verification. Not PRC-labeled. | Terms of Use §4.4 permit training on inputs and outputs for its own purposes. An enterprise no-training carve-out exists but is not verified for our account; no blanket API exemption is published. API output availability of at least seven days is not a deletion commitment. |
| Hangzhou DeepSeek Artificial Intelligence Co., Ltd. | Model inference, direct DeepSeek API (deepseek-chat, deepseek-reasoner) |
China (PRC) — data collected, processed, and stored on servers in the PRC | Privacy policy states inputs are used to train and improve its models and services |
| Moonshot AI (Kimi) | Model inference, Moonshot's own API (kimi-k3, moonshot-v1-*) |
China (PRC-headquartered). International platform terms state Singapore-located servers; endpoint serving our account pending verification | Platform privacy policy states user content is used to train and refine its models |
| NVIDIA Corporation (hosted NIM API catalog) Opt-in only — organization training consent required |
Model inference (Nemotron model family) — only via the separate, clearly-labeled NVIDIA-served options (nemotron-3-super-nvidia, nemotron-3-ultra-nvidia; a third option, nemotron-3-nano-nvidia, was retired by NVIDIA effective August 28, 2026), re-added behind the training consent gate on August 14, 2026. The default Nemotron models (nemotron-3-super, nemotron-3-nano, nemotron-3-ultra) route to DeepInfra under its no-training terms (see Subprocessors above) and never touch NVIDIA — between August 10 and August 14, 2026 no NVIDIA-served options existed at all, and no traffic reaches NVIDIA unless an administrator has opted the organization in. |
United States | Published API-catalog trial terms permit NVIDIA to use inputs/outputs to improve its products and services, including AI models — NVIDIA-served options are therefore locked by default for every organization and exist only behind an organization administrator's explicit training consent. |
Client control. If your organization does not want traffic reaching
independent-controller providers, contact us to pin your account's allowed models —
pinning is enforced at the gateway.
Other service providers
Data enrichment (Harmonic). The platform's business-data features use Harmonic, a company-data enrichment provider that receives only public company identifiers (such as company names and website domains) — never client request content, prompts, completions, or personal data of client users. A second enrichment provider, Apollo.io, is listed in the infrastructure & platform table above ahead of enablement; its row states its own, wider scope (identifiers of third parties a user asks about, and company domains) and it is not enabled.
Updates to this list
This list is maintained on the Platform per Annex 5 of our Data Processing Addendum and is reviewed periodically. When we engage a new subprocessor that will process Personal Data, we update this page and provide any notice required by the DPA. Items marked pending verification reflect characteristics we have not yet confirmed from official documentation — confirm before relying on them.
Questions about this list or our DPA: contact security@aireserve.com.