> ## Documentation Index
> Fetch the complete documentation index at: https://nova.dweet.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Legal Changelog

> Changes to Nova's Terms of Service and Privacy Policy

<Update label="April 2026" tags={["Privacy", "Sub-processors"]}>
  **Planned: April 15, 2026 | Effective: April 22, 2026**

  ### Unipile SAS added as sub-processor for outbound messaging and LinkedIn Recruiter

  We added Unipile SAS as a sub-processor for candidate outreach messaging and for LinkedIn Recruiter connection and shortlist export workflows.

  **What changed:**

  * Unipile SAS added to sub-processors list (section 3.8, Outreach, messaging, and LinkedIn Recruiter integration)
  * Used for sending and tracking candidate outreach messages via a unified messaging API
  * Used for LinkedIn Recruiter connection, verification, project listing, and shortlist export flows
  * Applies only to customers using outbound messaging and/or the LinkedIn Recruiter integration

  **What this means:**

  * Data sent to Unipile's API: candidate names and profile URLs (outbound messaging), outreach message content, LinkedIn session tokens (access token and premium token), browser user-agent, Recruiter contract identifier, and the IP address Nova's server derives from the connect request. For shortlist exports, Nova sends the Recruiter project identifier and the LinkedIn profile identifier (public identifier / Recruiter provider id) of each candidate being added. Nova's own export attempt history (success, partial success, or failure counts, and candidate-level attempt results) stays in Nova and is not sent to Unipile
  * Data processed by Nova (not forwarded to Unipile) from the connect flow: extension version, browser family, the paired Recruiter tab URL, Nova's detected LinkedIn surface and confidence score, cookie-presence flags, a short open-tab summary, and the Recruiter seat identifier used only as a transient safety signal during the connect flow (not persisted for later display), all used for pairing and safety checks before the session is imported
  * Data processed exclusively in France (European Union), hosted on Scaleway
  * Unipile is SOC 2 Type II certified and GDPR compliant
  * No international data transfers required (EEA processing only)
  * No impact if you don't use outbound messaging or LinkedIn Recruiter integration
  * If you have concerns, you may reply to the change notice or email `privacy@dweet.com` within the notice period set out in [DPA](/docs/nova/legal/dpa) Section 8.2

  The LinkedIn Recruiter connection uses a new browser extension, "Nova for LinkedIn," described in Section 11 of the [Privacy Notice](/docs/nova/legal/privacy-policy).

  [View Sub-processors →](/docs/nova/legal/subprocessors)
</Update>

<Update label="February 2026" tags={["Privacy", "Sub-processors"]}>
  **Planned: February 7, 2026 | Effective: February 14, 2026**

  ### Mistral AI added as sub-processor (Talent Pool only)

  We added Mistral AI as a sub-processor for AI-powered OCR to extract structured data from resume documents in Talent Pool features.

  **What changed:**

  * Mistral AI added to sub-processors list
  * Used for OCR-based resume extraction
  * Only applies to customers using Talent Pool features

  **What this means:**

  * Resume PDFs are processed by Mistral's OCR API (Talent Pool only)
  * Data processed in the European Union
  * Mistral does not use customer data for model training
  * No international data transfers required (EEA processing)
  * No impact if you do not use Talent Pool

  ### Fireworks AI removed for OCR, restored for AI inference

  We removed Fireworks AI, Inc. from resume extraction (now handled by Mistral AI) and re-added it for general AI inference, embeddings, and hosting of custom fine-tuned models.

  **What changed:**

  * Fireworks AI removed from resume extraction pipeline
  * Fireworks AI re-added for AI inference, embeddings, and fine-tuned model hosting
  * Inference data is not stored or logged; content exists only transiently during request processing

  ### Baseten Labs, Inc. added as sub-processor

  We added Baseten Labs, Inc. for AI inference, embeddings, and hosting of custom fine-tuned models.

  **What changed:**

  * Baseten added to sub-processors list
  * Processing in the United States and European Union
  * Inference data is not stored or logged; content exists only transiently during request processing

  ### Together AI, Inc. added as sub-processor

  We added Together AI, Inc. for AI inference, embeddings, and hosting of custom fine-tuned models.

  **What changed:**

  * Together AI added to sub-processors list
  * Processing in the United States and European Union (Sweden)
  * Inference data is not stored or logged; content exists only transiently during request processing

  ### OpenRouter, Inc. added as sub-processor

  We added OpenRouter, Inc. as an AI model routing layer used to access AI inference providers.

  **What changed:**

  * OpenRouter added to sub-processors list
  * Processing in the United States and European Union
  * Routes requests to underlying AI providers

  ### Modal removed as sub-processor

  We removed Modal.com from our sub-processors list as we no longer use their services for document conversion.

  **What changed:**

  * Modal removed from sub-processors list (previously section 3.8)
  * Document conversion functionality now handled differently

  **What this means for all changes above:**

  * All new AI inference sub-processors handle data transiently with no persistent storage of customer content
  * Standard GDPR transfer safeguards (EU SCCs, UK IDTA) apply for US processing
  * No sub-processor uses customer data for model training
  * If you have concerns about any new sub-processor, you may reply to the change notice or email `privacy@dweet.com` within the notice period set out in [DPA](/docs/nova/legal/dpa) Section 8.2, including your reasonable data protection grounds

  [View Sub-processors →](/docs/nova/legal/subprocessors)
</Update>

<Update label="December 2025" tags={["Privacy", "Sub-processors"]}>
  ### Fireworks AI added as sub-processor (Talent Pool only)

  We added Fireworks AI, Inc. as a sub-processor for hosting and running our custom fine-tuned AI model for resume extraction in Talent Pool features.

  **What changed:**

  * Fireworks AI added to sub-processors list
  * Hosts our custom fine-tuned vision-language model for resume extraction
  * Only applies to customers using Talent Pool features

  **What this means:**

  * Resume PDFs are processed by our custom model hosted on Fireworks (Talent Pool only)
  * Data processed in the United States
  * Fireworks does not use customer data for model training
  * Standard GDPR transfer safeguards apply
  * No impact if you do not use Talent Pool

  [View Sub-processors →](/docs/nova/legal/subprocessors)
</Update>

<Update label="January 2025" tags={["Privacy", "Analytics"]}>
  ### Analytics for authenticated users

  Product analytics is now enabled for all logged-in users to help us improve Nova, identify bugs, and understand feature usage.

  **What changed:**

  * Product analytics (PostHog) enabled for authenticated users
  * Privacy Policy Section 10 updated

  **What this means:**

  * We collect usage data (pages viewed, features used)
  * Data processed by PostHog in the EU
  * No third-party advertising sharing

  [View Privacy Policy →](/docs/nova/legal/privacy-policy)
</Update>